WEBVTT

1
00:00:00.359 --> 00:00:01.170
Josh Moore: kick that off.

2
00:00:02.490 --> 00:00:04.589
Josh Moore: So yeah Good morning, so i'm just more.

3
00:00:05.460 --> 00:00:06.899
Josh Moore: From the open microscopy environment.

4
00:00:07.890 --> 00:00:15.269
Josh Moore: Thanks everyone for coming and we'll do a round of introductions in just a second tried to keep it under 20 Minutes that we're not.

5
00:00:16.020 --> 00:00:30.120
Josh Moore: spinning too long on it um for anyone who's here for the first time, so to explain very briefly, this is generally a fairly technical conversation so you're welcome to kind of get into the details a bit.

6
00:00:30.660 --> 00:00:39.180
Josh Moore: We do only have two hours um with the goal of having a standard file format that we're all using in microscopy.

7
00:00:40.320 --> 00:00:49.110
Josh Moore: largely focused on this concept of n G F F, I mean f's did I say that I say enough so it's basically a cloud based file format that you can store in s3.

8
00:00:50.520 --> 00:01:03.720
Josh Moore: Many of the things we'll talk about, though, are also applicable to hdfs and to tiff and so, in general, we want to come up with a generalized solution, a small number of formats, that we can all use together.

9
00:01:04.110 --> 00:01:12.780
Josh Moore: And kind of in the madness that has developed over the past, you know 20 ish years, those are least the 20 that i've experienced and it goes back farther.

10
00:01:13.860 --> 00:01:18.720
Josh Moore: Where we've needed something like bio formats to actually access our data so that's the goal.

11
00:01:19.800 --> 00:01:24.270
Josh Moore: And the general focus of today will be trying to help.

12
00:01:25.320 --> 00:01:30.390
Josh Moore: Others, ie more of you to get involved in this.

13
00:01:32.010 --> 00:01:40.680
Josh Moore: Experience in this effort i'm add new specifications update the implementations in the various languages, the currently their.

14
00:01:41.880 --> 00:01:51.570
Josh Moore: Java Python and javascript implementations can we as a Community start working together to make that happened a bit faster, so this is a snowball snowball process.

15
00:01:53.010 --> 00:02:06.150
Josh Moore: um, I will start pointing at people i'll kind of go around my screen, I will warn in the chat who's up next for introducing themselves kick it off with Jason and then sets up that.

16
00:02:07.740 --> 00:02:12.150
Jason Swedlow: morning quickly i'm just as fellow at the University of Dundee.

17
00:02:13.440 --> 00:02:18.780
Jason Swedlow: And I work with josh and many of the others somewhere on the screen on the on the project.

18
00:02:25.260 --> 00:02:29.340
Josh Moore: Still around he turned off his video he may have kids so.

19
00:02:30.390 --> 00:02:31.500
Josh Moore: Go to Constantine then.

20
00:02:33.300 --> 00:02:48.180
Constantin Pape: Yes, hi and Constantine i'm a postdoc with an occasional Heidelberg and I work on vacation for large electron microscopy data until of this process i'm also interested in developing the five five minutes to share this.

21
00:02:55.830 --> 00:03:00.480
Deborah Schmidt: hi i'm Deborah i'm obese or something yeah and kind of having slept was also here.

22
00:03:01.740 --> 00:03:13.020
Deborah Schmidt: i'm currently working on side view, where we are thinking this 3D virtual reality view or we might want to be present a project, for example in our that's at least one idea.

23
00:03:13.800 --> 00:03:22.230
Deborah Schmidt: And so, by unprepared for this meeting, not exactly sure what will happen or what I can contribute but i'm looking at to listen and just get an idea of what's happening.

24
00:03:25.770 --> 00:03:41.400
Draga Doncila Pop: hi i'm draga i'm done on the student in while I just finished on those in at one atmosphere and i'm currently contracting for an apartment and basically here today to talk about I guess collections dimensions and resolve.

25
00:03:42.720 --> 00:03:47.070
Josh Moore: Oh yeah if you know the specification, you want to talk about do mention it excellent thanks trying.

26
00:03:52.380 --> 00:04:07.800
Guillaume Gay: i'm in charge of kind of standing Romero on the lumina campus in masa and i'm interested in bizarre because I think I face a lot of different.

27
00:04:08.850 --> 00:04:14.280
Guillaume Gay: Images a format and so on, so i'm happy to engage with the problem.

28
00:04:18.450 --> 00:04:28.290
Eric Perlman: americ problem and i'm at the Jackson laboratory actually have two hats that I care about this one is, I still do some contracting with dobby Bach whose lab I was in.

29
00:04:28.590 --> 00:04:39.690
Eric Perlman: For large scale em reconstructions which a bunch of you have here have actually worked on as well, but I also now at Jackson we have we use Romero, and so I care about how we can have.

30
00:04:41.040 --> 00:04:43.770
Eric Perlman: Access to very large slide scans and things of that sort.

31
00:04:49.140 --> 00:04:51.930
Josh Moore: Oh young years super quiet okay.

32
00:04:57.510 --> 00:05:00.750
Josh Moore: Maybe someone else or fix this quickly okay cool.

33
00:05:06.300 --> 00:05:07.530
Jean-Karim Heriche: Sorry, I forgot, I was muted.

34
00:05:08.760 --> 00:05:18.390
Jean-Karim Heriche: i'm shonky I am I work at end I do computational support for the cell biology and biophysics unit, so I encounter images a lot.

35
00:05:19.380 --> 00:05:31.080
Jean-Karim Heriche: In various forms and formats and i'm also work package leader in the ios LIFE project, so the European open science cloud.

36
00:05:31.860 --> 00:05:48.810
Jean-Karim Heriche: Implementation targeting the life sciences and for this I also we present your by Beijing in that project, and so we are particularly interested in how to deal with images in the cloud and particular image data access.

37
00:05:51.990 --> 00:05:53.310
Thanks Yun you want to try it again.

38
00:05:56.430 --> 00:05:57.180
Josh Moore: Yes, thank you.

39
00:05:57.660 --> 00:06:00.750
Jan Albrecht: Wonderful okay so i'm young i'm workouts.

40
00:06:02.430 --> 00:06:12.210
Jan Albrecht: And we're heavily interested in having a standard across the image field, because we want to build.

41
00:06:13.560 --> 00:06:18.150
Jan Albrecht: yeah methods to include all tools together and.

42
00:06:19.380 --> 00:06:20.550
Jan Albrecht: To provide some.

43
00:06:21.780 --> 00:06:24.630
Jan Albrecht: yeah layer to join all of them.

44
00:06:27.120 --> 00:06:27.840
Jan Albrecht: makes sense.

45
00:06:30.000 --> 00:06:32.010
Jon Fuller: hi i'm john fuller.

46
00:06:33.270 --> 00:06:43.710
Jon Fuller: i'm a scientific software engineer and have some experience in cloud I protecting and work for a dossier a smallish pharmaceutical company and bustle.

47
00:06:45.060 --> 00:06:51.630
Jon Fuller: also contribute to the orbit image analysis and open source software for whole site imaging.

48
00:06:52.770 --> 00:07:07.770
Jon Fuller: And at dorsey we're interested in primarily household site images and specifically today, I was interested in how to exchange annotations along with the image themselves.

49
00:07:10.590 --> 00:07:10.920
Jon Fuller: Thanks.

50
00:07:12.870 --> 00:07:28.200
Juan Nunez-Iglesias: So i'm one i'm a researcher at monash University in Melbourne and I work on the Perry and second image mostly and I would like to talk about physical spacing and physical position in the spec.

51
00:07:34.500 --> 00:07:48.930
Kyle Harrington: i'm kyle i'm a group leader at the Max kell brook Center in Berlin we're particularly interested so i'm also representing the helmholtz imaging platform so we're kind of a consortium of all the Humboldt instance in Germany.

52
00:07:50.310 --> 00:07:58.740
Kyle Harrington: We work on a bunch of scales, but here we're particularly interested in storing the storage of mesh formats, I was involved in that forum discussion a bit.

53
00:07:59.250 --> 00:08:10.770
Kyle Harrington: strings are storing storing Roi is where we're from the Fiji world, so we had been using and five, but we prefer to have the same standard everyone else's so we're happy with them fives are backing.

54
00:08:13.710 --> 00:08:24.270
Martin Schorb: I am Martin, I work at the electron microscopy facility at em bl and we're dealing with the different files for different formats different scales that can range from large tiles.

55
00:08:25.500 --> 00:08:41.220
Martin Schorb: Different modifications rotations light microscopy data for plan, but also very large volumes and I like to show all of that in context, so the spatial relation and orientation transformation registration that's my main interest.

56
00:08:44.790 --> 00:08:59.040
Nicolas Vinuesa: hi i'm good and they worked in a place where we are working with the pathology and microscopy data images actually and we are using by formats on video from recently we're also using by format true.

57
00:09:00.450 --> 00:09:10.680
Nicolas Vinuesa: So you're very interested in knowing the next file formats for cloud storage so yeah very happy to be here.

58
00:09:12.330 --> 00:09:12.930
Nicolas Vinuesa: Thanks for joining.

59
00:09:14.850 --> 00:09:27.030
Paolo Grisoli: hi my name is powder on Nicholas colleague at night, and yes, he said that we are starting using by farmers to grow in our tool chain and.

60
00:09:28.440 --> 00:09:34.860
Paolo Grisoli: We started within five so we look forward to see what's The next step in five four months to.

61
00:09:36.060 --> 00:09:37.440
Paolo Grisoli: To include it in our.

62
00:09:38.790 --> 00:09:40.410
Paolo Grisoli: In our implementation so.

63
00:09:42.330 --> 00:09:42.960
Paolo Grisoli: Nice to be here.

64
00:09:45.690 --> 00:09:52.560
William Moore: hi i'm will more work on me team mostly on a mirror and the website of things.

65
00:09:59.460 --> 00:10:00.960
Josh Moore: David and I think that's everyone.

66
00:10:04.710 --> 00:10:12.480
David Gault: yeah difficult part of the meeting, have mostly been working on the bio format side of their projects.

67
00:10:16.920 --> 00:10:21.210
Josh Moore: awesome so that was right at 10 minutes huge applause to everyone.

68
00:10:21.780 --> 00:10:36.420
Josh Moore: yeah so I mean Deborah mentioned not being prepared, so these really aren't like about having any prepared it's it's much more just getting everyone together to talk so do feel free to just kind of pipe up with whatever you need I was going to start with a.

69
00:10:37.710 --> 00:10:44.640
Josh Moore: kind of a status report from my side if anyone else has wants to do the same thing you can i'm sorry I should replace this.

70
00:10:45.570 --> 00:10:56.910
Josh Moore: The link to the to the notes i'll put it back in the chat i'm feel free to start writing things under the session one notes, if there are bits you want to get off your.

71
00:10:57.840 --> 00:11:10.020
Josh Moore: chest, there are links down at the bottom, so if you have projects that you want to link other people to throw it into that document, and then, when all of this is done and I managed to find time I will you know write it up on images see.

72
00:11:12.060 --> 00:11:22.860
Josh Moore: That you know eventually so status update So there are a couple of things that are going on taking a step back, especially for those who are just joining, for the first time i'm.

73
00:11:24.870 --> 00:11:30.360
Josh Moore: There were conversations I think that began after the I to K, the images to knowledge meeting.

74
00:11:32.100 --> 00:11:40.500
Josh Moore: um where there was kind of a clear direction that yes, there was interest in Czar and in five there wasn't a lot of interest in having more and more formats.

75
00:11:40.860 --> 00:11:50.370
Josh Moore: So we all decided we were going to try to find a way to work together right so that's where these calls began, and all the conversations on the on him and Jesse.

76
00:11:51.930 --> 00:11:53.940
Josh Moore: So most immediately.

77
00:11:55.920 --> 00:12:04.680
Josh Moore: The tsar project and only me projects received eos funding so essential open source software funding from Chen sucker Berg initiative.

78
00:12:05.250 --> 00:12:16.230
Josh Moore: i'm to basically push this work forward so the tsar work was specifically on i'll talk about that first was specifically on getting to the next specification v3.

79
00:12:17.310 --> 00:12:34.260
Josh Moore: um the goal of that was basically to unify the tsar and in five specs right, so they are both formats, are very, very similar very simple um and basically if we made some small adjustments, we can bring the two together so The hope is that, as our v3 the spec.

80
00:12:35.610 --> 00:12:47.670
Josh Moore: will be a proper subset of all the tsar work and all the end five work right so that's where we're going um the funding is coming to an end, it was a bit rough doing all this during the pandemic so.

81
00:12:48.840 --> 00:12:53.490
Josh Moore: The way it worked was the funding went to numb focus, so if anyone knows pi data.

82
00:12:54.690 --> 00:12:56.580
Josh Moore: numb focus is a project that helps.

83
00:12:58.020 --> 00:13:02.430
Josh Moore: Open Source projects take money and distribute them to contractors.

84
00:13:03.270 --> 00:13:11.130
Josh Moore: um we did that there are a couple of different contractors who came in and out so there wasn't like a nice continuation to get the project done.

85
00:13:11.730 --> 00:13:20.700
Josh Moore: um the state of where everything is is that there are branches of the specification the v3 specification their branches that.

86
00:13:21.390 --> 00:13:28.770
Josh Moore: Have the tsar Python project which can read the specification there's a c++ implementation um.

87
00:13:29.580 --> 00:13:36.060
Josh Moore: The implementations that haven't been done yet and may get done over the next couple of months, our Java Czar v3.

88
00:13:36.990 --> 00:13:46.050
Josh Moore: javascript and I think someone began a rust implementation, so this is something we will talk about throughout the rest of the next two hours.

89
00:13:46.500 --> 00:13:55.830
Josh Moore: um that, as we have more and more implementations of the specification, it becomes more and more difficult to keep things up to date so that's just a general problem that will have to think about.

90
00:13:57.420 --> 00:14:14.790
Josh Moore: um from the all me point of view um what we've been focusing on our so largely again for for everyone who's joined new leading up to the to the 2020 images to knowledge meeting in December so end of November, beginning of December.

91
00:14:15.900 --> 00:14:21.090
Josh Moore: We focused on storing what we call of labels I think that's what some of you will call annotations.

92
00:14:22.140 --> 00:14:25.230
Josh Moore: Some people will call mask but basically having a way to.

93
00:14:27.540 --> 00:14:33.300
Josh Moore: record a segmentation of your images, with some metadata and on storing high content screening.

94
00:14:36.000 --> 00:14:43.020
Josh Moore: Since then, largely in 2021 we've been focusing more on a 3D use case of those are primarily to the use cases.

95
00:14:45.900 --> 00:14:54.960
Josh Moore: have run into problems that I assume, others have run into when using Czar it's a problem that doesn't exist in the in five world which is.

96
00:14:55.320 --> 00:15:02.460
Josh Moore: All of the chunks are stored in the same directory so as you start to have more and more chunks in your file format eventually it's I think the largest I hit was.

97
00:15:03.210 --> 00:15:07.440
Josh Moore: 10s if not hundreds of millions of files in a single directory and things start to fall over.

98
00:15:08.250 --> 00:15:15.420
Josh Moore: So basically and that's a problem of the specification as well, so we're kind of taking a step back, trying to fix some of these underlying problems.

99
00:15:16.380 --> 00:15:27.660
Josh Moore: So we're not really actively involved with a specification right now i'm are basically trying to get the 3D performance up and do some benchmarking.

100
00:15:29.610 --> 00:15:47.670
Josh Moore: If there's a next specification that we would start to think about that would be Meta data, basically, there needs to be a breaking change from from oil, these are V1 to V2 or sorry the oh point one to Point two and that's probably will we're still we will start looking next.

101
00:15:48.720 --> 00:15:54.510
Josh Moore: But maybe the conversation that we're about to have will change that but that's just our general are rough roadmap.

102
00:15:55.170 --> 00:16:02.580
Josh Moore: um The other thing that's driving us is that more and more we're receiving data sets that can't easily be put into a marrow.

103
00:16:03.270 --> 00:16:09.090
Josh Moore: um and so even the idea, so the image data resource the idea.

104
00:16:09.690 --> 00:16:25.440
Josh Moore: has no marrow underneath it, and so, if you get to the point where there's data that doesn't work well in tomorrow, you need something else we assume that something else is only bizarre and we will start then trends forming converting users data into armies are for production.

105
00:16:27.090 --> 00:16:37.830
Josh Moore: which will also drive the need to have very clear milestones where you know we will support and oh point 1.2 or three because otherwise the data will get lost in it.

106
00:16:39.870 --> 00:16:43.380
Josh Moore: So those are kind of our drivers and needs that we need to see happen next.

107
00:16:44.790 --> 00:16:45.510
Josh Moore: um.

108
00:16:47.610 --> 00:16:55.860
Josh Moore: That went faster than I was expecting that's excellent, are there any questions or thoughts about that, so I went through it super fast, so if anything was unclear.

109
00:16:59.820 --> 00:17:02.010
Josh Moore: feel free to unmute and just scream.

110
00:17:06.090 --> 00:17:09.570
Juan Nunez-Iglesias: I think I don't want is meant to be broken so i'm happy with that.

111
00:17:13.020 --> 00:17:13.740
Josh Moore: it's good to hear.

112
00:17:14.190 --> 00:17:25.080
Constantin Pape: I mean, I have some thoughts on the label specification sort of when I saw the budget more, but I don't know you want to dive into the specifics now already.

113
00:17:25.680 --> 00:17:32.910
Josh Moore: Well let's let's see if there's anyone who's just kind of lost about what we're doing in general and labels is certainly on the list of specifications to spend some time on.

114
00:17:35.460 --> 00:17:38.880
Josh Moore: Okay, so does everyone have a general feel of what we're trying to accomplish here.

115
00:17:41.820 --> 00:17:47.970
Josh Moore: All right, and Is there anyone else from the Community who's been working on something that I didn't cover.

116
00:17:50.220 --> 00:17:52.320
Josh Moore: I try to keep abreast of what's going on but.

117
00:18:01.050 --> 00:18:05.700
Josh Moore: Okay, so I mean obviously feel free to throw things out there um.

118
00:18:07.350 --> 00:18:08.970
Josh Moore: And, and maybe we are getting to labels, then.

119
00:18:10.440 --> 00:18:25.680
Josh Moore: Constantine so there's a list of specifications listed in the document and they're all the ones I know of that someone has said they're interested in working on, so we can just go through them kind of top to bottom, spend some time on each of them to get people up to speed.

120
00:18:27.030 --> 00:18:35.010
Josh Moore: If there's a handful of people who want to work on like two or three separate ones, we can just break out and actually get real work done so that's always a good thing.

121
00:18:35.700 --> 00:18:44.490
Josh Moore: um and if there are more that people are thinking about and desperately need for what you know so looking at the list you know, I think.

122
00:18:45.810 --> 00:18:47.040
Josh Moore: we've got everything covered a.

123
00:18:48.120 --> 00:18:50.460
Josh Moore: lot I may be wrong, so.

124
00:18:53.820 --> 00:18:55.770
Josh Moore: I know I mentioned anything new, at least not yet.

125
00:19:00.900 --> 00:19:05.550
Josh Moore: Okay, so failing having other people want to talk let's just give the Florida Constantine.

126
00:19:06.960 --> 00:19:19.140
Josh Moore: What was that I put you know a good deal of time just to go through all the specifications and then keep in mind the real goal is to figure out who will take whatever we say and actually drive it forward so right that's what we're looking for.

127
00:19:20.400 --> 00:19:22.590
Juan Nunez-Iglesias: Right just intervene in the.

128
00:19:24.000 --> 00:19:28.320
Juan Nunez-Iglesias: In the army fyi says support an idea and hiring and it's that.

129
00:19:28.500 --> 00:19:29.610
Josh Moore: No, no, oh I want.

130
00:19:29.730 --> 00:19:31.620
Juan Nunez-Iglesias: To make one that you want to skip.

131
00:19:32.220 --> 00:19:41.010
Josh Moore: No, no, I didn't mean I didn't mean to skip it um yeah so we're in the process of hiring one position based in Dundee and.

132
00:19:41.580 --> 00:19:54.540
Josh Moore: Two three positions based at singer for this work and are basically you know, trying to drive it from our side, if there is anyone else who has either upcoming physicians or.

133
00:19:55.530 --> 00:20:02.850
Josh Moore: Someone who wants to you know get involved, we can do what it takes to have I don't know if it's a slack or a julep or.

134
00:20:03.660 --> 00:20:11.130
Josh Moore: integrate people into these efforts, so we will, you know as the people finally show up these new developers.

135
00:20:11.580 --> 00:20:24.210
Josh Moore: will hopefully make them known to everyone else in the Community and then make it clear this person is working on, you know X i'm and try to drive those you know, in the back of my mind, I wonder.

136
00:20:25.230 --> 00:20:30.930
Josh Moore: Is that a working group, you know how formalized we want to do this, you know so far this has been very.

137
00:20:33.210 --> 00:20:41.190
Josh Moore: relaxed and ad hoc, you know as long as we're making progress, I think that's fine I have been asked at a couple of meetings.

138
00:20:42.480 --> 00:20:54.660
Josh Moore: You know, does is this work following some some process, you know there's a process from the rba the research data alliance, you know, or is it following.

139
00:20:55.200 --> 00:21:09.660
Josh Moore: I triple E or is you know we're not following anything, yet which is OK, but you know if if people feel they're not finding a way to get involved in these this conversation, then please raise a flag or yell at us, or you know.

140
00:21:10.680 --> 00:21:12.750
Josh Moore: We try to be you're raising your hand Jason.

141
00:21:14.250 --> 00:21:14.550
Jason Swedlow: Just.

142
00:21:16.650 --> 00:21:29.820
Jason Swedlow: The I guess the implication of what josh was saying is obviously from an odd perspective we're taking this very seriously at in the process of so less you have any doubt.

143
00:21:30.810 --> 00:21:45.090
Jason Swedlow: We don't think we have the resources currently to deliver what the Community needs, I mean full stop, we do not, the challenge is too great so we're in the process of submitting more proposals.

144
00:21:48.210 --> 00:21:57.270
Jason Swedlow: to grow, the team, and I just say from you know, representing all me all of those are constructed as partnerships with.

145
00:21:57.870 --> 00:22:04.890
Jason Swedlow: partners, in particular, you will you sort of heard what josh was implying, for example, working with sanger.

146
00:22:05.490 --> 00:22:15.300
Jason Swedlow: The ones we're working on now are actually international partnerships so looking for partnerships across boundaries, so the extent to which that is interesting to any of you.

147
00:22:15.840 --> 00:22:30.450
Jason Swedlow: And it's something that you or your groups or institutions, whatever are interested in driving, please do make contact, because it would be great to do that and to do that in a way, where we have.

148
00:22:32.970 --> 00:22:41.340
Jason Swedlow: I was gonna say as formal as possible and I, you know that gets into what josh was just talking about, but at least the glue between us to get this work done.

149
00:22:42.360 --> 00:22:49.140
Jason Swedlow: You know, at least in terms of the teams and the constructs how the mechanisms as josh was saying, would be a completely different story so.

150
00:22:50.250 --> 00:22:53.280
Jason Swedlow: that's probably my two cents, I will shut up and get out of the way but.

151
00:22:55.890 --> 00:23:12.390
Jason Swedlow: Definitely, please do get in touch with you have some interest in doing that, and you, especially if you know some mechanism to get some a reasonable amount of funding and I without casting aspersions anywhere anywhere reasonable means at least a couple years and at least a couple sours.

152
00:23:13.680 --> 00:23:15.420
Jason Swedlow: Sorry, just one.

153
00:23:16.620 --> 00:23:17.850
Josh Moore: More funding source Nice.

154
00:23:18.840 --> 00:23:22.770
Jean-Karim Heriche: All right, yep I was going to say something.

155
00:23:24.090 --> 00:23:27.300
Jean-Karim Heriche: along the lines that maybe i've been.

156
00:23:28.500 --> 00:23:36.300
Jean-Karim Heriche: Somewhat semi formal process might actually help in recruiting more for more.

157
00:23:37.560 --> 00:23:43.290
Jean-Karim Heriche: Contributions i'm thinking, maybe something like you're the lead with some of the.

158
00:23:44.940 --> 00:23:51.180
Jean-Karim Heriche: Software open source software project, maybe on our rfc based type of.

159
00:23:52.200 --> 00:23:54.030
Jean-Karim Heriche: of things and.

160
00:23:55.200 --> 00:24:13.590
Jean-Karim Heriche: From my point of view, I think it might help in an onboarding some of the vendors like microscopy companies so that actually they they have some people for for implementing.

161
00:24:14.190 --> 00:24:24.180
Jean-Karim Heriche: file formats, so if they could actually devote some of that default away from their own proprietary format to work, something that.

162
00:24:25.200 --> 00:24:27.420
Jean-Karim Heriche: is more Community based.

163
00:24:28.500 --> 00:24:36.420
Jean-Karim Heriche: But I think it would help to have at least some sort of formal formal process or semi formal process for that.

164
00:24:37.560 --> 00:24:39.120
Jean-Karim Heriche: And i'm also wondering.

165
00:24:41.280 --> 00:24:46.080
Jean-Karim Heriche: If you could involve maybe your by imaging in here.

166
00:24:48.900 --> 00:24:55.920
Jean-Karim Heriche: Maybe to mediate with the the companies, I don't know Jason probably knows more than me yeah.

167
00:24:56.700 --> 00:25:01.560
Josh Moore: Well, interestingly i'm john Federico may join in the afternoon session so let's.

168
00:25:01.590 --> 00:25:05.580
Jean-Karim Heriche: OK so maybe it was asking them what they think about that.

169
00:25:05.820 --> 00:25:08.340
Josh Moore: i'll place that question to them when they went in if they come.

170
00:25:10.590 --> 00:25:19.860
Constantin Pape: From MIT of the year, like partially funded by zero imaging to do a bit of graphic might also yeah touch touch the specs.

171
00:25:24.900 --> 00:25:25.650
Josh Moore: Constantine gets.

172
00:25:26.730 --> 00:25:34.470
Josh Moore: The pole position but yeah others welcome and just thinking about the process Sean cream so we'll.

173
00:25:35.190 --> 00:25:41.670
Josh Moore: do some screen sharing, by the time we're done here and least kind of show what I would think would be a good way so using the NGA FF.

174
00:25:42.540 --> 00:25:57.990
Josh Moore: repository but open to suggestions, so it might be that we can take the outcome of what we discussed and people who want to sign up for things you know assign a sign Constantine a task if that's what he's if he's on board and then others can can jump in.

175
00:26:00.060 --> 00:26:01.620
Josh Moore: anyone else on kind of the.

176
00:26:01.980 --> 00:26:03.510
Josh Moore: High Level funding fun.

177
00:26:05.190 --> 00:26:06.120
Josh Moore: process yep.

178
00:26:06.390 --> 00:26:08.640
Jon Fuller: So just one thing from my site and.

179
00:26:10.530 --> 00:26:23.220
Jon Fuller: Maybe coming to what john cream was mentioning about bringing vendors on board, I think one thing that we've noticed it also that it's possible is if you're buying an instrument, then of course you have quite a lot of.

180
00:26:25.530 --> 00:26:34.200
Jon Fuller: ways to motivate vendors to to contribute to these kind of things so that's a requirement on your side and you're about to give a vendor a lot of money, then that's.

181
00:26:35.580 --> 00:26:45.930
Jon Fuller: that's an easy way to contribute to push vendors in that direction would be something that we've seen in the past, at least on our side, maybe, if anyone else is.

182
00:26:46.980 --> 00:26:50.160
Jon Fuller: kind of binding instruments than that something just to bear in the back of your mind.

183
00:26:51.690 --> 00:26:52.080
Josh Moore: yeah.

184
00:26:52.320 --> 00:26:56.520
Josh Moore: Agreed, I mean we certainly have been telling people that for years, but I guess the question is what.

185
00:26:57.060 --> 00:27:03.900
Josh Moore: What can we do so, what can this group do to help people when they're purchasing so i'm not exactly sure what that is you know.

186
00:27:05.940 --> 00:27:12.420
Josh Moore: A formal statement of support for what we're doing, and you know the the vendors must sign it i'm not sure, but ideas welcome.

187
00:27:13.320 --> 00:27:15.300
Jon Fuller: I mean, I guess the the.

188
00:27:16.590 --> 00:27:24.180
Jon Fuller: practical thing on the side of the person purchasing is you can say you do you support this file format, or can you support this file format.

189
00:27:25.500 --> 00:27:29.730
Jon Fuller: it's being the first step to that and then maybe having.

190
00:27:30.900 --> 00:27:36.780
Jon Fuller: You know from there that's like the the sort of stick side of things, and the carrot might be something.

191
00:27:37.920 --> 00:27:41.760
Jon Fuller: Some statement that would help a vendor to kind of say okay well how do I.

192
00:27:43.950 --> 00:27:45.690
Jon Fuller: kind of make those steps to support that.

193
00:27:50.910 --> 00:27:59.220
Jean-Karim Heriche: I think the one of the issue is that it depends a little bit on the process by which the microscopes are required.

194
00:28:00.240 --> 00:28:16.440
Jean-Karim Heriche: So, for example, I don't have any say in in in the decision for for the microscope I can basically say well ask them if they support days but that's basically right at the bottom of the list of things that.

195
00:28:17.790 --> 00:28:19.170
Jean-Karim Heriche: People would want so.

196
00:28:20.460 --> 00:28:32.430
Jean-Karim Heriche: yeah I don't know if you're a facility had and you or someone in charge of making the decision, yes, but otherwise I think it's difficult and and also, I think it's.

197
00:28:33.750 --> 00:28:50.040
Jean-Karim Heriche: probably better to have something that already exist when you ask for for features for for availability, so I think might be too premature here, but on the other, and maybe involving the the companies into.

198
00:28:51.690 --> 00:28:56.430
Jean-Karim Heriche: Contributing so that they actually feel that they are needs also addressed.

199
00:29:01.110 --> 00:29:01.590
Josh Moore: Definitely.

200
00:29:04.410 --> 00:29:05.940
Josh Moore: will get there, one way or the other, so.

201
00:29:07.260 --> 00:29:08.880
Josh Moore: More resources, the faster we'll get there.

202
00:29:10.170 --> 00:29:13.590
Josh Moore: Okay, so shall we start talking about details.

203
00:29:14.850 --> 00:29:18.510
Josh Moore: Because that's why many of us are here Constantine us start with labels.

204
00:29:20.010 --> 00:29:37.350
Constantin Pape: I mean okay from from my side, so there are two points I mean the first one, which might be a bit discussion so just I mean i've been looking at the whole idea again and i'm still not sure how happy I am with this everything living sort of under one big container.

205
00:29:38.610 --> 00:29:45.180
Constantin Pape: Where the issue sort of, in particular for kind of Labor or segmentation for like very big volume.

206
00:29:45.840 --> 00:29:54.000
Constantin Pape: em a test right, you will only want to ship sort of the labels when they updated you don't want to ship, the whole raw data again.

207
00:29:54.690 --> 00:30:03.750
Constantin Pape: And of course it's still feasible right if everything is sort of under one container, but in that case at least some of their needs to be some.

208
00:30:04.350 --> 00:30:14.640
Constantin Pape: tools to help the someone who wants to do this, so that you can really just ship to the relevant part and also update the parts and the Meta data.

209
00:30:15.150 --> 00:30:20.310
Constantin Pape: That do this I think if someone is just sort of presented with this folder structure and this.

210
00:30:20.940 --> 00:30:30.540
Constantin Pape: That death, the raw data, and then we have these 10 different segmentation in these other sub folders and now at another one, this might be sort of a bitch.

211
00:30:31.410 --> 00:30:44.910
Constantin Pape: yeah not ideal, from your perspective and mad people just might be people to just say oh i'm gonna ship, the whole thing again or just just not feasible if you have like a way to put out by passion to say well data set.

212
00:30:46.200 --> 00:30:55.950
Josh Moore: So from I can't remember it was the the December color the October call, and we certainly added a issue here i'll put in the notes it's remote remote.

213
00:30:56.610 --> 00:31:03.030
Josh Moore: um So for me it that falls under the concept of supporting remote links in the tsar.

214
00:31:03.690 --> 00:31:14.280
Josh Moore: um yeah, so I think we I think everyone was completely on board and my reasoning, for not having done it was just starting off simple and getting more complicated, so I.

215
00:31:15.090 --> 00:31:20.760
Josh Moore: OK, so I think that means there's nothing needs to change on the labels implementation from what you're saying.

216
00:31:21.480 --> 00:31:27.180
Josh Moore: But you would need to be able to follow a link to a remote data set so you would have you know the original data volume.

217
00:31:27.690 --> 00:31:42.060
Josh Moore: Well, actually, it probably goes the other way around, you could put as a segmentation and Azhar somewhere think so, and then that could link to the original data, and so, if you Double Click you know you try to view the segmentation you'll you'll find a way to get access to the.

218
00:31:42.780 --> 00:31:44.790
Josh Moore: Data right exactly yeah.

219
00:31:45.930 --> 00:31:52.140
Constantin Pape: yeah yeah I just think it was great sort of to put this into production would be pretty important.

220
00:31:53.070 --> 00:32:10.980
Constantin Pape: And then the second part would be for for the version opi to to think about specifications for Meta data like especially transformations which is something we are interested in sort of this large data your data view our mobi site.

221
00:32:12.780 --> 00:32:18.780
Constantin Pape: And they I think they are, I mean two proposals that i'm around one is this open organization.

222
00:32:20.670 --> 00:32:24.360
Constantin Pape: Data layout and then we run, we did for one of the.

223
00:32:25.440 --> 00:32:37.800
Constantin Pape: workshops for the new bias and yeah sort of their work, to the question like, how do we actually go into the process of making a proposal to to add these things to spec.

224
00:32:40.590 --> 00:32:44.190
Constantin Pape: And if there are other competing proposals out there, we should be around.

225
00:32:46.530 --> 00:32:51.150
Josh Moore: yeah let's actually leave so that's probably the most interesting question.

226
00:32:52.380 --> 00:32:57.510
Josh Moore: So does anyone else have specifications that we should compare.

227
00:33:00.420 --> 00:33:00.840
Jon Fuller: Something.

228
00:33:00.930 --> 00:33:02.340
Josh Moore: else already done this work, so I go for it.

229
00:33:03.870 --> 00:33:06.600
Jon Fuller: I spent a little bit of time and we've had a quick.

230
00:33:07.830 --> 00:33:14.760
Jon Fuller: conversation in the form with bank it on I assume that when we talk to you about labels, I mean this is.

231
00:33:15.540 --> 00:33:28.410
Jon Fuller: In the case of whole site imaging i'm thinking about a as a pathologist draw something on an image and I want to or there's an algorithm specifies something on an image and it's essentially a Polygon or a rectangle.

232
00:33:29.670 --> 00:33:30.540
Jon Fuller: That, I want to them.

233
00:33:31.890 --> 00:33:41.010
Jon Fuller: That has some metadata attached to that as well you know, maybe a label of this is a whatever organ organ l.

234
00:33:42.210 --> 00:33:44.820
Jon Fuller: or cell and I want to share.

235
00:33:46.740 --> 00:33:50.160
Jon Fuller: with other software so.

236
00:33:51.390 --> 00:33:57.480
Jon Fuller: This this probably a couple of things use cases that we kind of head on our site that we thought about so one is.

237
00:33:59.160 --> 00:34:07.620
Jon Fuller: That you know I I annotated something in orbit and I want to annotate that in order to view those annotations in queue path.

238
00:34:08.970 --> 00:34:11.760
Jon Fuller: Maybe there's some different functionality, you want to between the two.

239
00:34:12.840 --> 00:34:29.310
Jon Fuller: And then second use case it's really going down these kind of I think is quite typical into the deep learning cases where you say okay Now I want to output these labels to some arbitrary algorithm running in a container maybe some black box, essentially, I want to output some.

240
00:34:30.390 --> 00:34:44.970
Jon Fuller: images and some labels and I want to then returned some labels back and view them on my image in that case, and I think for sure transformations is something that you want to be able to support in depth.

241
00:34:46.080 --> 00:35:02.880
Jon Fuller: But then it's also kind of thinking, a bit about whether you want to also namespace certain bits as well, so you say well you know I I have something that actually it's probably only relevant to orbit but I namespace that and if someone else wants to.

242
00:35:04.980 --> 00:35:16.080
Jon Fuller: You know use that part of my my annotation then they can, but then they're kind of down to my aka that I talked to find in the on the orbit namespace essentially.

243
00:35:17.880 --> 00:35:19.080
Jon Fuller: So that was kind of what was.

244
00:35:20.160 --> 00:35:22.290
Jon Fuller: done to kind of use cases a bit on our side.

245
00:35:24.570 --> 00:35:29.370
Jon Fuller: Maybe I let that sink I also had a question, because my ignorance.

246
00:35:30.540 --> 00:35:37.110
Jon Fuller: And so on this outside if I store as our file and on s3.

247
00:35:38.430 --> 00:35:47.250
Jon Fuller: Does that exist in as a single object, or is that that's multiple objects, but just within a defined format okay great Thank you.

248
00:35:51.090 --> 00:35:53.130
Josh Moore: anyone else have to add to that.

249
00:35:53.580 --> 00:36:05.580
Deborah Schmidt: I have a question, yes, yes, what john just talked about is vector based labeling right and some of us are dealing with takes a base level links, I guess, and.

250
00:36:06.750 --> 00:36:24.840
Deborah Schmidt: That could be a spouse or then like so different modalities and i'm not sure exactly what the current proposal is about, and I was also wondering where I can look it up just to get some stuff about my disk image slivers version 0.1 yes okay.

251
00:36:25.020 --> 00:36:35.520
Josh Moore: The current yeah and it's um you know, so the the extent of what's there is in the latest version of the NGA FF spec.

252
00:36:36.360 --> 00:36:47.160
Josh Moore: So there's but I don't think there's anything new that's happened that's not on that one thread Deborah so just for everyone's benefit the current extent of the label spec is.

253
00:36:49.080 --> 00:36:58.290
Josh Moore: um let's take a step back, so the first back was basically how to specify and multi scale image right, so you want to have a.

254
00:36:59.490 --> 00:37:11.310
Josh Moore: If if the image so coming from the army world all the images are currently five D, so one asks under the tbd section do they have to be five D, currently, yes, but that's certainly something we can talk about.

255
00:37:13.080 --> 00:37:16.170
Josh Moore: um and then every five D volume.

256
00:37:17.190 --> 00:37:26.820
Josh Moore: can have a some multiple arrays a pyramid of multi scale arrays for having the sub resolutions, so you know that's just an image.

257
00:37:27.630 --> 00:37:31.740
Josh Moore: The only thing that the label spec does is it says Okay, you can have an image.

258
00:37:32.430 --> 00:37:39.990
Josh Moore: With it sub resolution, and you can define that it's a segmentation it's a it's a labeled image and it can point back then to the original data.

259
00:37:40.590 --> 00:37:56.760
Josh Moore: That currently is within the same Czar so you can have the labeled image and the original data, you have a pointer from one to the other so that's what I was talking about a second goal is eventually we want to have remote pointers and you can specify so color metadata and.

260
00:37:57.810 --> 00:37:59.280
Josh Moore: annotations per.

261
00:38:00.360 --> 00:38:11.460
Josh Moore: pixel value in the labeled in and the labeled image right and that's kind of it, for the moment right so to the extent that we want to do more with the pixel based.

262
00:38:11.940 --> 00:38:27.150
Josh Moore: labeling we can add to that specification and then there's been several conversations around the polygons, and so we could actually maybe shift to polygons next on to kind of keep those two conversations close together and I guess meshes kind of it is involved um.

263
00:38:28.410 --> 00:38:32.550
Josh Moore: You know, so what is that going to look like, so I think.

264
00:38:37.800 --> 00:38:48.480
Josh Moore: There have been a small handful of proposals on the the Polygon representation so that's well known text well known binary and I forgotten the one that Anatole nested it's.

265
00:38:52.260 --> 00:38:57.300
Josh Moore: it's basically a format that also allows you to store meshes would kind of bring polygons and meshes together.

266
00:38:58.440 --> 00:39:04.680
Josh Moore: So I do think we want to support the two worlds and then it's kind of a question of just getting all the features that everyone wants from from each of them.

267
00:39:06.840 --> 00:39:07.380
Josh Moore: Go for mine.

268
00:39:08.370 --> 00:39:16.860
Martin Schorb: What about sub volumes and individual object so, especially in the cryo vm world where you identify individual proteins and a volume.

269
00:39:18.120 --> 00:39:27.030
Martin Schorb: You want to just give the position of each rhizome and maybe an orientation or a state as well, and that would be the information that should also somehow.

270
00:39:28.170 --> 00:39:28.590
Martin Schorb: fit.

271
00:39:31.200 --> 00:39:38.580
Martin Schorb: So it's basically a list of objects with you know some some metadata for each object like orientation state.

272
00:39:40.380 --> 00:39:54.600
Martin Schorb: and probably a sub volume associated somewhere else, where you can then load the atomic structure of this thing in position, so you don't need to fill the entire volume with this structure but it's like very sparse but very high detail.

273
00:39:55.950 --> 00:39:57.570
Josh Moore: I mean I wouldn't say no for.

274
00:39:58.020 --> 00:40:05.100
Juan Nunez-Iglesias: Some volume, do you mean pixels do you mean like a little hyper cube R cube or.

275
00:40:05.400 --> 00:40:06.510
Martin Schorb: It can also be I.

276
00:40:06.510 --> 00:40:18.450
Martin Schorb: Would you like to spend literally also be a vector eyes model of the of the molecule like where each atom sit like the chemistry bonds in in some spectra specification.

277
00:40:19.020 --> 00:40:21.030
Martin Schorb: it's not necessarily volume data.

278
00:40:21.330 --> 00:40:21.540
that's.

279
00:40:24.090 --> 00:40:26.910
Kyle Harrington: The point clouds are meshes representation, I think.

280
00:40:29.040 --> 00:40:36.810
Martin Schorb: In some sense, yes, but that's only it's only very sparse very little objects here in their place in the whole volume.

281
00:40:38.970 --> 00:40:53.400
Josh Moore: I mean, I guess, for me, so and I may be just you know holy naive, so I guess, this is one of the things I wanted to say up front is that you know I consider you guys to be the ones who are saying what actually is valuable and needs to be defined.

282
00:40:54.810 --> 00:41:02.910
Josh Moore: And it's my job, just to like get you to come to a consensus and specify it and support it for a long time right so um.

283
00:41:05.310 --> 00:41:12.840
Josh Moore: What that sound like sounds like, though, to me, is is part of the Polygon specification for, for example, just being able to have a table of points.

284
00:41:13.680 --> 00:41:24.030
Josh Moore: In in higher dimensional space and you were defining things and associating metadata with them, you know, there will, can be a table of data for every point or for a set of points or or right um.

285
00:41:25.290 --> 00:41:29.430
Josh Moore: I think that's consistent but someone would have to tell me if that doesn't meet all the requirements.

286
00:41:34.140 --> 00:41:34.800
Guillaume Gay: If I may.

287
00:41:36.660 --> 00:41:36.990
Guillaume Gay: I.

288
00:41:37.320 --> 00:41:38.610
Josh Moore: Just do it, I can see you.

289
00:41:39.150 --> 00:41:40.980
Guillaume Gay: know my webcam is out sorry.

290
00:41:42.480 --> 00:41:48.180
Guillaume Gay: So I looked a bit into polygons and Mrs this last since the last meeting.

291
00:41:50.070 --> 00:42:09.720
Guillaume Gay: I think, having both in the same world is difficult because meshes are have a kind of strict definition and meshes imagine, people are expecting to see faces and vertices and that set and adding to it means bending the specifications of it, I think.

292
00:42:11.160 --> 00:42:23.520
Guillaume Gay: But I think what you said that the use case where you have a a protein at a certain place, I think, a point plus Meta data pointing to ptb entry, for example.

293
00:42:24.240 --> 00:42:43.860
Guillaume Gay: And then just having the whatever representation, you want from TV, I think that this in polygons basically arrow eyes or put polygons on our eyes, yes, and I think, as long as your console points and points Meta data to that location, I think it's it's better to split the two.

294
00:42:45.600 --> 00:42:46.260
Guillaume Gay: The two worlds.

295
00:42:48.360 --> 00:42:54.360
Guillaume Gay: So yeah that's my two cents, as for as long as then, then they can live the mind as for meshes.

296
00:42:55.380 --> 00:43:00.060
Guillaume Gay: I think, p Li can very easily be written in za.

297
00:43:01.350 --> 00:43:09.060
Guillaume Gay: And is a well known specification for meshes and it's really a no brainer to write a PA lies our.

298
00:43:11.190 --> 00:43:13.740
Guillaume Gay: code to have meshes just like.

299
00:43:14.970 --> 00:43:17.070
Guillaume Gay: easy and understandable missions for everyone.

300
00:43:18.750 --> 00:43:28.200
Josh Moore: that's it for tonight thanks good young, can you give us drop a URL into the document for what you think would be easy to represent thanks john Kerry.

301
00:43:29.670 --> 00:43:40.290
Jean-Karim Heriche: was going to say something a little bit along the same line but also considering point clouds So the question would be if we consider point to be.

302
00:43:41.160 --> 00:43:52.710
Jean-Karim Heriche: The minimal Polygon if you want, because what I think also Martin was alluding to reset for some some points you would also associate a vector basically.

303
00:43:53.370 --> 00:44:05.310
Jean-Karim Heriche: have, for example, indicating directions, or that could be indicating value so The other thing, so it could be basically numerical make Meta data on the audience.

304
00:44:08.670 --> 00:44:18.810
Jean-Karim Heriche: Then, so what talking about the pi format reminded me that I don't know we haven't followed necessarily all the discussion, but if you guys have looked at the vr l.

305
00:44:20.040 --> 00:44:21.030
Jean-Karim Heriche: format also.

306
00:44:25.980 --> 00:44:38.190
Jean-Karim Heriche: just another way also of representing any eight is basically specification for four points and polygons and I think meshes yes i've only looked at it, a long time ago, but.

307
00:44:39.900 --> 00:44:43.590
Guillaume Gay: to remember that i'm not sure it's like i'll go ahead.

308
00:44:44.310 --> 00:44:55.620
Kyle Harrington: So so i'm pretty certain that poi it really actually covers all of these cases, because you can extend our betray attributes you don't have to even have faces.

309
00:44:55.980 --> 00:45:02.550
Kyle Harrington: In all cases, it could just be only described vertices it can describe arbitrary numbers of dimensions.

310
00:45:03.180 --> 00:45:08.610
Kyle Harrington: So you can, if you want to talk to attach orientations to every point that that's no problem.

311
00:45:08.970 --> 00:45:22.290
Kyle Harrington: I mean, one of the reasons but it actually gave him and I were talking about the discussions you can store chemical kinetic information at each point anything you want it's really just a table at the end of the day.

312
00:45:24.720 --> 00:45:30.270
Guillaume Gay: I had questions about hold because you can't stop you can define a hole.

313
00:45:31.560 --> 00:45:35.370
Guillaume Gay: In pri or so they say in the specification so.

314
00:45:36.990 --> 00:45:40.560
Guillaume Gay: That that was maybe you don't have that notion of yeah yeah.

315
00:45:41.580 --> 00:45:43.770
Guillaume Gay: So maybe for for very generic.

316
00:45:44.850 --> 00:45:47.130
Guillaume Gay: Li it's not adapted I don't.

317
00:45:48.690 --> 00:45:59.370
Eric Perlman: link into another topic we're gonna hit here, what are the things i'm curious with the meshes is going to be multi scale and also fragments so we imagine very large meshes that could be.

318
00:46:00.240 --> 00:46:16.470
Eric Perlman: Millions or even like a billion faces, how do we represent, that in this format so tied into the Multi scale so like we have meshes each of the Multi scale and then do we chunk them along the same chunks as the boundary where is this going to be sort of a parallel chunky.

319
00:46:18.210 --> 00:46:19.380
Josh Moore: I have seen some.

320
00:46:20.910 --> 00:46:33.960
Josh Moore: i'm trying to remember where I saw it i'll find the link and send around, certainly if we kick off the meshes conversation, again, I will add it there um Maybe it was in geosciences someone was looking at multi scale meshes.

321
00:46:35.160 --> 00:46:38.730
Josh Moore: explicitly Eric so yeah So hopefully someone else has solved it but.

322
00:46:39.330 --> 00:46:50.820
Eric Perlman: that's an Arab lands or has a very custom multi scale mesh implementation, which is something we could use as maybe a reference not for the format per se, but as to like what is involved.

323
00:46:53.760 --> 00:47:01.830
Constantin Pape: Just off of the hard things from this i've had left, right and para who also have a pretty custom way to do with the motif specialist.

324
00:47:02.730 --> 00:47:03.090
Okay.

325
00:47:06.600 --> 00:47:18.450
Eric Perlman: yeah exactly between the between those we can figure out which features are necessary, I mean it sort of comes down to defining the scales and then defining how each of those scales are chunked in the end dimensional space that you have.

326
00:47:22.980 --> 00:47:33.510
Eric Perlman: The hard part becomes being being the one who's actually rendering then and choosing which one to access but by having the combination there you actually get you know fast ui and interaction to these things.

327
00:47:41.790 --> 00:47:48.060
Jean-Karim Heriche: And I know the soda, that is, that I was thinking is maybe if there's.

328
00:47:48.540 --> 00:48:06.480
Jean-Karim Heriche: Too much trouble representing many things in one format is using linking to specific representation so again, for example, taking the example of associating a vector to a point, or you don't necessarily have to have that vector in the same way presentation, but you could.

329
00:48:08.160 --> 00:48:14.400
Jean-Karim Heriche: link to let's say a table or to something so treat some of that information, maybe as.

330
00:48:15.420 --> 00:48:28.470
Jean-Karim Heriche: Separate I mean it's more maybe more leaking maybe maybe at some point, it comes with a penalty but in terms of performance but might make things easier to to split things a little bit more.

331
00:48:30.570 --> 00:48:31.980
Josh Moore: yeah there's definitely a.

332
00:48:38.640 --> 00:48:44.490
Josh Moore: The costs, you know will start to grow exponentially if we try to do everything in one format, so I.

333
00:48:46.950 --> 00:48:58.680
Josh Moore: i'm hoping that others will, for example on for tape tabular tabular data there's something called the awkward arrays project right in Python and they're working together with is our team.

334
00:48:59.160 --> 00:49:11.280
Josh Moore: You know, and it allows you to store pretty complicated tabular data in bizarre you know so to the extent that someone else comes up with a way to represent complicated.

335
00:49:12.450 --> 00:49:17.370
Josh Moore: data and metadata and Czar we can make use of that but I agree on agreement you know.

336
00:49:17.850 --> 00:49:30.750
Josh Moore: If we have to bend over backwards to represent something or we have to redo it entire spec, then we should stop and and reevaluating right, you know we shouldn't redo all of vr ml, for example, or all of anything.

337
00:49:31.680 --> 00:49:47.850
Josh Moore: I know gilliam I can't remember who you got in touch with, but you know asked can this mesh format be stored in Czar and the the developer basically said, why would you do that, you know that doesn't make sense, so there's certainly a a fine line that we have to keep in mind.

338
00:49:50.250 --> 00:49:51.660
Guillaume Gay: That was the mesh io.

339
00:49:52.290 --> 00:49:53.400
Josh Moore: misha misha.

340
00:49:54.240 --> 00:49:54.960
Josh Moore: cheers Thank you.

341
00:49:55.800 --> 00:50:00.360
Josh Moore: um okay so aware of the time, and I mean we can always look back to these.

342
00:50:01.380 --> 00:50:15.810
Josh Moore: Two kind of summarizing questions for labels polygons and meshes does anyone else have a what was the word pretty custom implementation that they know up that they want to get on the list of things that should be evaluated on don't feel embarrassed.

343
00:50:17.280 --> 00:50:17.700
Josh Moore: You know.

344
00:50:20.730 --> 00:50:27.060
Josh Moore: Research software engineers at rest develop new formats that's just what happens um and to.

345
00:50:27.090 --> 00:50:30.660
Josh Moore: Is there anyone who wants to you know put their name beside.

346
00:50:32.460 --> 00:50:35.760
Josh Moore: These items and keep working on it after this meeting is over.

347
00:50:37.230 --> 00:50:40.350
Josh Moore: So first any more implementations that anyone wants to throw up.

348
00:50:40.860 --> 00:50:47.790
Jon Fuller: So on the Polygon sage m P threw up to GEO json.

349
00:50:49.230 --> 00:50:51.300
Jon Fuller: spec I can just post a.

350
00:50:52.920 --> 00:51:01.500
Jon Fuller: comment after send the link here, but there was there was what we found very practical and there's a bunch of implementations in.

351
00:51:03.300 --> 00:51:05.190
Jon Fuller: several languages, which is kind of handy.

352
00:51:09.300 --> 00:51:14.940
Jon Fuller: I know, at least on the Polygon side of things, I would be happy to continue to be involved in.

353
00:51:17.910 --> 00:51:33.180
Eric Perlman: Something I would like to see on the Polygon side would be useful tools for extracting the associated bite values from Azhar re assuming that we are using polygons for our allies associated with it image volume.

354
00:51:38.190 --> 00:51:38.940
Juan Nunez-Iglesias: That sounds fun.

355
00:51:41.370 --> 00:51:42.960
Eric Perlman: polygons to didn't ask type in.

356
00:51:44.790 --> 00:51:47.520
Juan Nunez-Iglesias: The risk of turning my old on a student under a bus.

357
00:51:50.130 --> 00:52:07.290
Juan Nunez-Iglesias: This is this is more related to the collections thing it's still pretty custom, but but there's some good ideas in this format, but dragons come up, which is you know it's a collection of images each image has a segmentation and it would be good to be able to open.

358
00:52:08.640 --> 00:52:17.820
Juan Nunez-Iglesias: Individual images with their Center segmentation or the whole collection has a thing so it's a bit of a hierarchical thing it's pretty custom.

359
00:52:18.870 --> 00:52:24.840
Juan Nunez-Iglesias: But it's Nice, and it would be good to be able to do that sort of thing with armies are.

360
00:52:25.980 --> 00:52:31.290
Juan Nunez-Iglesias: Moving forward and and you know I think dragon looked at the whole plate specification and whatever.

361
00:52:33.270 --> 00:52:40.320
Juan Nunez-Iglesias: And it's very heavy weight, for if you don't actually have a plate right if you have some totally unrelated collection of stuff.

362
00:52:41.550 --> 00:52:46.020
Juan Nunez-Iglesias: Just like you know images have been finance right that's one of the things.

363
00:52:49.410 --> 00:52:49.740
Juan Nunez-Iglesias: yeah.

364
00:52:52.620 --> 00:52:53.640
Josh Moore: i'm going to.

365
00:52:55.650 --> 00:53:07.410
Josh Moore: Ignore you for a second one, so I mean that's a good point and I actually have a few things to add to that, but did not get us completely into collections until we've signed off on labels meshes and polygons just very briefly okay.

366
00:53:07.620 --> 00:53:12.720
Josh Moore: Last call for implementations anyone else who wants to get involved young are you.

367
00:53:12.960 --> 00:53:16.530
Josh Moore: Still in, or are you tired of it yet.

368
00:53:19.590 --> 00:53:21.390
Guillaume Gay: No, no, I mean I mean okay.

369
00:53:22.050 --> 00:53:31.710
Kyle Harrington: So we were interested, but I actually the question I would have is what's going on with the Java SAR side of things is we're in Java land.

370
00:53:33.630 --> 00:53:42.630
Josh Moore: So I mean david's here and can give us more details there from our side, we will work on as our reader and writer.

371
00:53:44.850 --> 00:53:52.950
Josh Moore: That means that bio formats, will be able to read and write um I think we can then bootstrap up higher separately.

372
00:53:53.970 --> 00:53:56.280
Josh Moore: In bill Heidelberg has the movie.

373
00:53:57.300 --> 00:54:13.320
Josh Moore: You were which has its own implementations of Azhar reader they can also read from s3 and that's the big data viewer space i'd really like to kind of combine more of those things so that we're each depending on each other as opposed to everyone's forced to develop their own.

374
00:54:13.890 --> 00:54:23.250
Constantin Pape: And if I read some of our in our movie sinks, this also inherits from this outfit and five za.

375
00:54:23.460 --> 00:54:23.730
Okay.

376
00:54:25.110 --> 00:54:27.780
Kyle Harrington: that's that's our land as well, so yeah.

377
00:54:27.870 --> 00:54:44.160
Constantin Pape: I mean, I think this is fairly competitive but it's just sort of tiny things that needed to be duplicated and following horrible engineering practices is just duplicated code, but if someone was much knowledge would look at it as a respected.

378
00:54:46.260 --> 00:54:58.350
Josh Moore: For what it's worth so I was going to say this, towards the end tomorrow so every two weeks there's as Oracle is our Community call, as opposed to an old movies or Community call right it gets confusing um.

379
00:54:59.100 --> 00:55:11.670
Josh Moore: that's tomorrow night at eight o'clock Central European time sorry I do your own math um I guess for kyle it's the same time zone so it's fine um.

380
00:55:12.630 --> 00:55:26.970
Josh Moore: I have tried to get more of the Java Czar in five implemented years to come, so there are three that I know of, so the low level api's that's in five jays are and net CDF Java.

381
00:55:27.360 --> 00:55:32.940
Josh Moore: Hopefully they'll all be there and then we can start at the bottom to build you know it properly engineered layered.

382
00:55:33.990 --> 00:55:39.300
Josh Moore: code stack so if anyone would like to join you're welcome to that's tomorrow night yes.

383
00:55:40.320 --> 00:55:46.920
Josh Moore: um okay so but going back to the most important part of that statement kyle I put your name I marked.

384
00:55:48.060 --> 00:55:49.650
Josh Moore: you're interested that's good to know.

385
00:55:49.920 --> 00:55:52.590
Josh Moore: yeah all right anyone else.

386
00:55:53.310 --> 00:56:03.660
Deborah Schmidt: I will just quickly mention that I do have some experience with syfy also have this like I didn't do much there but there's anything I can help with on ease of your stuff.

387
00:56:08.100 --> 00:56:13.890
Deborah Schmidt: I just know that inside we we get a bunch of issues because of Sophia or he plugins right.

388
00:56:15.420 --> 00:56:16.170
Deborah Schmidt: I would really like.

389
00:56:21.750 --> 00:56:24.870
Josh Moore: yeah I mean it's it's going to be interesting, so at the moment.

390
00:56:26.010 --> 00:56:30.450
Josh Moore: You know, you can imagine, when we get to the point that more and more data is you know.

391
00:56:31.050 --> 00:56:37.830
Josh Moore: Okay, we convince the vendors and more and more data is being produced in this format all of our tools are producing the format, you know.

392
00:56:38.730 --> 00:56:46.230
Josh Moore: Things will have to keep getting easier because I do have the feeling we're all kind of struggling to work with yet another format it's the you know the X case ED.

393
00:56:46.680 --> 00:57:06.510
Josh Moore: 15th standards problem at the moment, you know completely realized that um so anything we can do to make our lives easier, but you know, step by step, big thumbs up so um so do scream if you're if you're kind of suffering and we'll try to figure out a way to fix it.

394
00:57:08.100 --> 00:57:10.140
Josh Moore: Okay anything else on a Polygon yep.

395
00:57:10.200 --> 00:57:16.800
Constantin Pape: I do we also continue involvement with like the not and fixing white labeling spec right now.

396
00:57:17.730 --> 00:57:18.180
Josh Moore: sure.

397
00:57:18.360 --> 00:57:25.170
Constantin Pape: I read this is what I would be interested in looking at the sort of getting to transformations in the next iteration.

398
00:57:25.710 --> 00:57:27.630
Josh Moore: Okay, so everyone so far was on.

399
00:57:29.040 --> 00:57:30.210
Josh Moore: vector based.

400
00:57:34.500 --> 00:57:38.310
Josh Moore: anyone else, particularly involved in want to be involved in in pixel based.

401
00:57:45.510 --> 00:57:47.700
Deborah Schmidt: I think that's where I went to us to okay right.

402
00:57:48.480 --> 00:57:56.400
Kyle Harrington: I mean in particularly the transform the pixel wise, yes, the transformations are also particularly interesting because we have those associated with each of our measures.

403
00:57:58.170 --> 00:58:05.400
Deborah Schmidt: And what I didn't get completely from the post was to you already support mighty live like mighty.

404
00:58:06.510 --> 00:58:08.400
Deborah Schmidt: mighty labels for pixel.

405
00:58:09.180 --> 00:58:09.420
Josh Moore: For.

406
00:58:09.480 --> 00:58:12.360
Josh Moore: pixel based labels, yes, definitely like.

407
00:58:12.420 --> 00:58:13.980
Deborah Schmidt: overlapping labels that works.

408
00:58:14.520 --> 00:58:15.120
So.

409
00:58:16.260 --> 00:58:32.220
Josh Moore: that's certainly something that can be discussed, you know can be improved, currently there are two ways to deal with overlapping labels one is um you can have a collection of labels, so you can say that the Union of all these labels represents what we want.

410
00:58:33.690 --> 00:58:35.250
Josh Moore: or specify.

411
00:58:36.300 --> 00:58:42.270
Josh Moore: i'm an extra pixel value for where the overlaps are right, you can you can define.

412
00:58:43.620 --> 00:58:45.960
Josh Moore: A, B and then the overlap is see right.

413
00:58:46.740 --> 00:58:51.900
Deborah Schmidt: yeah because this was a big thing I can my for my lip you clip that we often discuss because we have.

414
00:58:52.290 --> 00:59:03.900
Deborah Schmidt: These overlapping labels or the time and you cannot be presented in in one image, without any extra metadata like you can have a specific pixel value for the overlap, but you need to snap to.

415
00:59:04.860 --> 00:59:15.510
Deborah Schmidt: This overlap, we presents these two labels, or something like that right and we didn't know how to be present, that in a file format that others would also be able to return right.

416
00:59:16.170 --> 00:59:23.880
Josh Moore: yeah so you know going for it would certainly be good to get your data into the format and see which the two works best for you so.

417
00:59:24.420 --> 00:59:31.530
Josh Moore: we're possible if we don't have to support multiple solutions for these problems, you know someone was asking about the minimal specification so.

418
00:59:31.950 --> 00:59:46.500
Josh Moore: Try to keep the fewest number of things on the fewest number of of Sub specifications and then, if we need to add more in the future, we will like you know, for example, I don't think we want poi and vr ml, so we should choose like if we're using one of those do we should choose.

419
00:59:46.500 --> 00:59:49.560
Josh Moore: One and go as far as we can with it.

420
00:59:49.620 --> 00:59:56.280
Josh Moore: We may reach a point where we need to support multiple things but it's it complicates your your code, which is what we're trying to prevent.

421
00:59:57.150 --> 01:00:14.340
Josh Moore: An example of that is an image is always multi scale, even if there's just one resolution just so that you don't need an if statement every time you're reading the image right, so the extent that we can just do things the in a uniform way I think it'll make our life simpler.

422
01:00:15.630 --> 01:00:16.230
Josh Moore: But we'll get there.

423
01:00:17.910 --> 01:00:22.500
Josh Moore: Okay transitioning to collections on the bit I wanted to add.

424
01:00:23.880 --> 01:00:35.790
Josh Moore: One and drag is that I know, so I assume you read on the thread I think he'll be here in the afternoon, which isn't very useful for you.

425
01:00:36.300 --> 01:00:47.160
Josh Moore: i'm Christian Fisher, and I assume Martin you're involved in this i'm are interested in a collection specifications, so I would assume they would add their name um.

426
01:00:47.850 --> 01:00:56.490
Josh Moore: I i'm assuming that quite hardly and I will push um So if you have something you want others to look at that would be a good next person to look at it.

427
01:00:57.150 --> 01:01:11.160
Josh Moore: um and I think the collection specification, is one that can be done pretty independently, so I don't see any reason why it can't get in super fast like you did your label specification So for me the floor is open.

428
01:01:12.420 --> 01:01:22.350
Josh Moore: doors open windows open, you might just see if if you can get feedback from anyone else who's here on on feature requests that you know that need to be integrated.

429
01:01:23.730 --> 01:01:33.630
Josh Moore: So everyone, I mean one kind of went over what collections are about its bags of images and with whatever metadata you're interested in so.

430
01:01:34.800 --> 01:01:35.280
Josh Moore: Anyone.

431
01:01:39.180 --> 01:01:53.670
Martin Schorb: I assume you would generalize that and would also include meshes finery labels also as part of a collection, because that would then go to what what Constantine was asking that you kind of modular eyes things, and then only update.

432
01:01:54.750 --> 01:02:00.750
Martin Schorb: Like one part of one module of this like one new segmentation, for example, without touching the rest.

433
01:02:02.850 --> 01:02:03.870
Josh Moore: yeah I mean.

434
01:02:07.050 --> 01:02:09.060
Josh Moore: I don't see any reason not to do that so.

435
01:02:09.840 --> 01:02:24.060
Josh Moore: I don't know enough about what the mesh you know, whoever does the mesh specification I guess we'll see but currently labels are images therefore from in my thinking, if you have a collection of images, yes, it could definitely be a collection of labels right.

436
01:02:26.580 --> 01:02:34.830
Martin Schorb: yeah, the question is, if you do if you do collection of entities and entities will also include measures like a level above basically.

437
01:02:37.200 --> 01:02:39.360
Martin Schorb: I don't I don't see a conceptual difference there.

438
01:02:44.070 --> 01:02:57.690
Juan Nunez-Iglesias: yeah I think for me to really nice feature in in tragus pseudo spec is is the ability to open individual things as if they were the whole thing.

439
01:02:59.370 --> 01:03:03.570
Juan Nunez-Iglesias: And I would hope that whatever spec we come up with has that property.

440
01:03:05.190 --> 01:03:06.600
Constantin Pape: Do we do we have a link.

441
01:03:06.630 --> 01:03:08.250
Draga Doncila Pop: To that proposal somewhere.

442
01:03:10.050 --> 01:03:17.520
Draga Doncila Pop: So I can definitely add a link to these notes that's not a way it's you know custom is definitely a stretch.

443
01:03:17.520 --> 01:03:17.880
Constantin Pape: For it.

444
01:03:17.910 --> 01:03:22.140
Draga Doncila Pop: it's basically something that arose out of a need to.

445
01:03:23.070 --> 01:03:38.880
Draga Doncila Pop: support both opening an associated sort of you know bag as a good word for images for labeling but training purposes, but also then wanting to maintain a clear correspondence between the individual images and their individual segmentation.

446
01:03:39.420 --> 01:03:52.980
Draga Doncila Pop: I guess a future inspection and also for future you know editing so I guess that sort of speaks to what you were talking about as well about the modularity and whether like one big container in sufficient.

447
01:03:54.090 --> 01:03:58.950
Draga Doncila Pop: For example, when we were working on this and the power of one of the issues was that once you opened.

448
01:03:59.640 --> 01:04:15.630
Draga Doncila Pop: up a folder of images as a stack then they just sort of State A stack and so you know, to save them out into individual images again you sort of had to you know slice through that stack and ideally you'd be able to as one mentioned have a hierarchy what you can open.

449
01:04:16.770 --> 01:04:21.150
Draga Doncila Pop: The stack itself or just individual images and.

450
01:04:22.170 --> 01:04:41.430
Draga Doncila Pop: yeah I don't see any reason why it should only be supported by images, I definitely think that the collection of entities makes sense, definitely when I was working on this, like quasi spec the idea of having like points layers associated with images also seem to make perfect sense.

451
01:04:43.050 --> 01:04:46.470
Draga Doncila Pop: And then I guess in a sort of tangentially related discussion.

452
01:04:49.260 --> 01:04:54.750
Draga Doncila Pop: about the like spec being lightweight bottom up, and the five the hard coded requirements.

453
01:04:56.640 --> 01:05:05.520
Draga Doncila Pop: It would definitely be nice to be able to abstract a lot of the I guess metadata that's not needed away because.

454
01:05:06.480 --> 01:05:22.020
Draga Doncila Pop: What we found especially watching last year as well with remote sensing data is that Oh, and he was almost absolutely perfect to store these satellite images, but just not quite, and so there probably is I guess room for.

455
01:05:24.300 --> 01:05:24.900
Draga Doncila Pop: I guess.

456
01:05:26.820 --> 01:05:36.210
Draga Doncila Pop: I know amy's aspect that's further up in the hierarchy and defines even more of a bare minimum than what it currently does.

457
01:05:37.530 --> 01:05:47.130
Juan Nunez-Iglesias: If that makes sense, I think one of the things that drug and I discussed earlier today is, you know if you go to do means our spec now like the draft page.

458
01:05:48.330 --> 01:05:53.670
Juan Nunez-Iglesias: there's just so much stuff there, and so, if you're someone who's like trying to figure out.

459
01:05:55.140 --> 01:06:07.020
Juan Nunez-Iglesias: Can I store my data this way you'll just come here and you'll go like Well, this is not for me you wouldn't say, can I can I shoehorn my data into this i'm going to do singleton dimensions for time and.

460
01:06:09.720 --> 01:06:10.110
Josh Moore: So.

461
01:06:10.170 --> 01:06:11.190
Juan Nunez-Iglesias: which might be yeah.

462
01:06:13.200 --> 01:06:16.110
Josh Moore: i'm laughing because this is, I feel like a.

463
01:06:17.310 --> 01:06:21.060
Josh Moore: redheaded stepchild i'm sorry Jason i'm.

464
01:06:26.760 --> 01:06:27.270
Josh Moore: So.

465
01:06:28.470 --> 01:06:29.190
Josh Moore: i'm.

466
01:06:30.420 --> 01:06:36.960
Josh Moore: So, going back to when we first introduced are needed the Multi scale representation, so I went to this our Community.

467
01:06:37.800 --> 01:06:49.230
Josh Moore: Open up issues said, you know I really want to you know have this extension to what you're doing just want to be able to represent, you know multi scale data on kind of went back and forth get input from the Community and, eventually, they said yeah.

468
01:06:50.370 --> 01:06:59.400
Josh Moore: This is, this is, this is an art thing go, you know go do this in your own space I go over here I create the olympians our implementation, I do the Multi scale specification.

469
01:07:00.510 --> 01:07:04.770
Josh Moore: And now on say Oh well, we don't want it here send it back the other.

470
01:07:04.800 --> 01:07:06.600
Juan Nunez-Iglesias: No, no, no, no, not the Multi scale is fine.

471
01:07:06.600 --> 01:07:07.080
Sebastien Besson: Actually.

472
01:07:07.350 --> 01:07:09.240
Juan Nunez-Iglesias: It wasn't all right thing.

473
01:07:09.840 --> 01:07:18.000
Josh Moore: yeah, but I think what you're saying so So how do we solve it So what does the naming look like you know I think what you're actually saying is you want.

474
01:07:18.720 --> 01:07:28.710
Josh Moore: You know you're interested in a specification that lives within gfs as opposed to with the oil me metadata because I think that will me is going to continue to be more.

475
01:07:29.910 --> 01:07:31.770
Josh Moore: My car microscopy based.

476
01:07:33.240 --> 01:07:34.260
Josh Moore: We need to find a whole bunch.

477
01:07:34.260 --> 01:07:35.400
Josh Moore: Of the other things.

478
01:07:35.430 --> 01:07:46.470
Juan Nunez-Iglesias: Yes, I would like an NGO that has you know, because the thing is it doesn't have like a physical size, so it would be nice to have and I don't think there is even scope women for that.

479
01:07:48.000 --> 01:07:49.170
Juan Nunez-Iglesias: And so it would be nice if.

480
01:07:50.190 --> 01:07:57.210
Juan Nunez-Iglesias: The metadata can be defined minimally enough that people know.

481
01:07:58.500 --> 01:08:10.710
Juan Nunez-Iglesias: You know how to specify an image with the scale without having to deal with five dimensions or things like that, and how to specify collection, without having to deal with plates and wells, and things like that.

482
01:08:14.220 --> 01:08:14.580
Juan Nunez-Iglesias: yeah.

483
01:08:16.920 --> 01:08:27.510
Juan Nunez-Iglesias: And I am draga do you have that folder structure handy, would you be able to share that because I thought that was very compelling earlier.

484
01:08:31.920 --> 01:08:32.220
Juan Nunez-Iglesias: yeah.

485
01:08:34.140 --> 01:08:35.520
Juan Nunez-Iglesias: Is that is that.

486
01:08:36.180 --> 01:08:36.690
Jon Fuller: Maybe was.

487
01:08:36.750 --> 01:08:48.900
Jon Fuller: was a struggle sits at the screen share just to check I understood that I think I can probably have a similar opinion and the idea would be of, from my perspective would be that if you have a.

488
01:08:50.220 --> 01:08:57.390
Jon Fuller: Whole side image which is probably two dimensional and that you only want to actually care about those two dimensions.

489
01:08:58.050 --> 01:08:59.850
Jon Fuller: And you don't want to care about the extra stuff.

490
01:09:00.240 --> 01:09:04.230
Juan Nunez-Iglesias: You don't want to worry about i'm going to have a singleton time dimension, you want to just.

491
01:09:04.890 --> 01:09:09.780
Jon Fuller: have your kid just to be done for you yeah I think I would probably appreciate that.

492
01:09:11.010 --> 01:09:11.490
Josh Moore: So I think.

493
01:09:11.820 --> 01:09:12.390
Josh Moore: For me, that.

494
01:09:12.420 --> 01:09:21.450
Josh Moore: That that is part of all of these are you know we don't need to find a new home for that that just needs to happen, you know it's a um.

495
01:09:23.100 --> 01:09:28.860
Josh Moore: I don't even think it's a breaking change because it just kind of loosens what the clients can expect, so you know.

496
01:09:30.270 --> 01:09:40.650
Josh Moore: Again, going back to how do we make this practically happen, you know, starting with fewer assumptions means you can just get your client code working and get things out into the into the Community but.

497
01:09:40.890 --> 01:09:45.210
Josh Moore: Eventually, we need to support more of these use cases and you know 2d is definitely one of them.

498
01:09:46.710 --> 01:10:02.820
Josh Moore: You know, finding another home for something like the entire hcm or the you know the very kind of particular use cases um have to do some thinking about what the naming there is, but the the feedback has definitely been taken on board sorry go for it try.

499
01:10:04.110 --> 01:10:10.560
Draga Doncila Pop: yeah so the you know basic ideas you might have input data that just looks like a bunch of.

500
01:10:11.760 --> 01:10:25.080
Draga Doncila Pop: You know, leaves, for example, and you might want to annotate these now if you can open these as a stack obviously and the I guess desired behavior that we would have is, you would have a za.

501
01:10:26.670 --> 01:10:38.370
Draga Doncila Pop: Bear with the specification businesses in its infancy, but the idea, rather than being multi scale levels, the numbers here correspond to the slice in the stack that.

502
01:10:39.300 --> 01:10:51.210
Draga Doncila Pop: You were labeling and and inside each slice is all of the you know images labels I guess points meshes whatever else might be associated with a particular image.

503
01:10:52.620 --> 01:10:56.940
Draga Doncila Pop: At that slice level and the hierarchical aspect is that you can.

504
01:10:58.260 --> 01:10:58.860
classic.

505
01:11:07.770 --> 01:11:19.710
Draga Doncila Pop: said so, the idea is that you would be able to open the labels Liz and the image stack itself, but then you're also able to go in here and open just one individual slice.

506
01:11:20.340 --> 01:11:28.200
Draga Doncila Pop: You know, make some annotations to that save that out, or you can even go further, within and then just opened me a one particular image.

507
01:11:29.310 --> 01:11:40.080
Draga Doncila Pop: Or the associated label in a slice and so that's the hierarchical structure and i'm not you know I don't i'm not a user necessarily of I guess.

508
01:11:40.890 --> 01:11:52.620
Draga Doncila Pop: These tools and i'm not particularly familiar with you know microscopy or many of the I guess formats that are out there, but this seems widely enough applicable.

509
01:11:54.000 --> 01:11:59.610
Draga Doncila Pop: That it's something that we probably would want to support, at least in yelling in some variety.

510
01:12:06.180 --> 01:12:21.420
William Moore: that's essentially what we did with the current ACS spec right, and then you can go into the you can bounce down through your folder hierarchy down to the well and open an individual image and the labels that are associated with it will be.

511
01:12:23.220 --> 01:12:25.440
William Moore: shown you so you can open just one in each.

512
01:12:26.520 --> 01:12:27.000
William Moore: At a time.

513
01:12:28.410 --> 01:12:29.580
Juan Nunez-Iglesias: That angels will say.

514
01:12:29.700 --> 01:12:31.620
Juan Nunez-Iglesias: there's a there's a hierarchy which is.

515
01:12:31.650 --> 01:12:34.740
Juan Nunez-Iglesias: plates and wells is that is that right, and then.

516
01:12:36.030 --> 01:12:42.180
Juan Nunez-Iglesias: I guess yeah what I would say, is to make that option on I that's probably in your roadmap i'm not sure.

517
01:12:42.600 --> 01:12:44.910
Josh Moore: Well yeah I mean I think what we want is actually.

518
01:12:45.990 --> 01:13:00.720
Josh Moore: I mean, we would have done it, but if dragons doing it all the better, you know take whatever specification and then back port forward port, you know refactor the hdfs back to make use of the collection, you know so that's it.

519
01:13:00.870 --> 01:13:19.020
Sebastien Besson: It should be particular implementation of whatever you're doing it's a super type thing that corresponds to specific acquisition modality, but as far as I can tell there's there's nothing in what dragon just presented that fundamentally conceptually clashes with what he does.

520
01:13:20.400 --> 01:13:21.660
Juan Nunez-Iglesias: yeah awesome.

521
01:13:22.470 --> 01:13:34.590
Josh Moore: yeah and you know if someone so another one that is needed, certainly is the on the tissue use case, so you have a large issue you have images that are laid out, and you know kind of a very.

522
01:13:35.670 --> 01:13:39.660
Josh Moore: um exactly an orderly so someone just sent me tma.

523
01:13:41.130 --> 01:13:48.930
Josh Moore: Thank you, someone said that publicly um you know representation and each of those images has a transform to get everything.

524
01:13:49.260 --> 01:14:00.900
Josh Moore: represented properly, so you know if if that's what someone is already doing themselves, then it would be good to make sure that their collection representation works with draw draw goes works with the hts one.

525
01:14:01.500 --> 01:14:10.860
Josh Moore: And we'll just kind of keep adding these probably the same thing for the em use case that Martin was talking about it all just comes down to you know the same kind of stuff.

526
01:14:16.650 --> 01:14:16.950
Juan Nunez-Iglesias: cool.

527
01:14:17.520 --> 01:14:18.720
Josh Moore: Anyone yet go for it.

528
01:14:18.930 --> 01:14:19.890
Juan Nunez-Iglesias: Oh, that was just cool.

529
01:14:22.110 --> 01:14:31.560
Josh Moore: Has anyone else been doing this in the background secretly well, is it pretty custom implementation, they want to talk about sorry bespoke know we can you get a terminology, right here.

530
01:14:37.170 --> 01:14:37.710
Josh Moore: Go for it man.

531
01:14:37.770 --> 01:14:48.030
Martin Schorb: i'm not sure if I got that right, so what What you said is that you don't want to have this scale draga sorry that the scale shouldn't be the at the top level because of.

532
01:14:49.260 --> 01:14:52.440
Martin Schorb: comprehensiveness of the of the hierarchy.

533
01:14:53.580 --> 01:14:54.210
Martin Schorb: Did I get it.

534
01:14:56.010 --> 01:15:04.380
Draga Doncila Pop: Oh yeah, no, no, no, no, so sorry i'm definitely happy with the Multi scale like specification, you know exactly the way it is the.

535
01:15:07.260 --> 01:15:16.650
Draga Doncila Pop: The difference there was just in this particular implementation for the purposes of the reader and writer plugin which particular user asked for.

536
01:15:17.910 --> 01:15:27.660
Draga Doncila Pop: Those numbers represent the slice of image or more correctly an individual image which would be part of this new collection specification.

537
01:15:28.020 --> 01:15:38.070
Draga Doncila Pop: And I don't see any reason why they would have to be numbered I mean i'm sure that we could you know, come up with a way to name them more usefully or you know.

538
01:15:39.420 --> 01:15:41.580
Draga Doncila Pop: I don't know some sort of a zero dot zero.

539
01:15:43.500 --> 01:15:49.350
Draga Doncila Pop: So the situation, but I definitely have an issue with the Multi scale implementation, the way it is.

540
01:15:50.130 --> 01:15:56.970
Juan Nunez-Iglesias: Now, at the bottom of that hierarchy, you can have the Multi scale, with the single scale ish that that makes sense to me anyway.

541
01:15:57.690 --> 01:16:02.550
Draga Doncila Pop: or the other way around, I suppose, whichever way we decided this morning.

542
01:16:02.820 --> 01:16:11.340
Juan Nunez-Iglesias: The other way would be a bit trickier to because you wouldn't be able to have a single image right if you if you have the scales.

543
01:16:13.980 --> 01:16:21.810
Martin Schorb: Do you think that you could just define that in the collection definition at the very top like what layer represents what.

544
01:16:27.180 --> 01:16:29.520
Martin Schorb: The most flexible that in the collection definition.

545
01:16:29.520 --> 01:16:30.210
Martin Schorb: You define.

546
01:16:31.590 --> 01:16:35.280
Martin Schorb: know how your how your folder hierarchy what what that represents.

547
01:16:35.790 --> 01:16:43.380
Juan Nunez-Iglesias: yeah but the point is, if you have a multi scale image like one of the design requirements is to be able to open a single image and that collection.

548
01:16:45.270 --> 01:16:48.180
Juan Nunez-Iglesias: And so you can't put the Multi scale information at the top.

549
01:16:48.930 --> 01:16:59.730
Josh Moore: So i'm not several you fast enough to bring up or will bring up a the metadata for the hcl specification the lock that points to the images.

550
01:17:00.000 --> 01:17:04.800
Sebastien Besson: That was figured out that we had during the Ah yes, surely had a lot of.

551
01:17:06.330 --> 01:17:14.910
Sebastien Besson: A lot of back and forth from executive is problem of where the metadata should live, it should be at the top level at the at the bottom level and.

552
01:17:15.330 --> 01:17:23.910
Sebastien Besson: Well, surprisingly you're fine good argument for both which is down to exactly what dragon want, I want to image and an image by myself well so that data should get.

553
01:17:24.810 --> 01:17:32.310
Sebastien Besson: in depth and or I want to open for collection, so I actually want to have as much into detail the top level as possible, so.

554
01:17:32.850 --> 01:17:46.260
Sebastien Besson: it's a tricky balance and just is showing you and will take over of defining where the Meta data leads that contains everything and very suddenly the step of being able to reconcile all the metadata at the different levels.

555
01:17:47.040 --> 01:18:06.540
Josh Moore: So we went back and forth, I think we did three or four different versions, where there was one version, where all the metadata like a a json block was at the top level with a link to every image and that kind of got unwieldy, you know so that it's definitely a balance between.

556
01:18:08.340 --> 01:18:19.650
Josh Moore: You know, taking all the metadata that you're talking about Martin your where you you open up the collection, you know everything and that that the user would need to click twice to get somewhere.

557
01:18:20.160 --> 01:18:29.100
Josh Moore: um we chose for the https, to say the user needs to click twice to get down to an individual image um but that may be just a special.

558
01:18:29.550 --> 01:18:39.120
Josh Moore: You know implementation detail of hdfs because there are so many layers of the hierarchy, you know, in a standard collection I assume you will want.

559
01:18:39.810 --> 01:18:50.520
Josh Moore: You know, and this is the type of block that we may want to standardize you will want a a top level thing that says here's how you get to the things that are multi scales right.

560
01:18:52.200 --> 01:18:55.560
Josh Moore: um and then you know if a user clicks on that.

561
01:18:56.880 --> 01:19:01.320
Josh Moore: Because what you what I think what we found and will can correct me if you don't want.

562
01:19:01.770 --> 01:19:17.580
Josh Moore: The client to need to go and do you know lots of look ups to figure out what images to show the user you kind of want to go tell me what i'm opening opening you know and everything else, at least when you start to get into large numbers becomes very, very painful.

563
01:19:23.550 --> 01:19:26.040
Jean-Karim Heriche: So, seems to me that.

564
01:19:27.930 --> 01:19:38.400
Jean-Karim Heriche: Maybe maybe i'm wrong, but this discussion reminds me of some discussions i've had about designing your database and how to relate the different items and.

565
01:19:39.420 --> 01:19:44.250
Jean-Karim Heriche: And, basically, the way it works is that you basically build.

566
01:19:47.130 --> 01:20:04.140
Jean-Karim Heriche: relational tables that relates your your indices, so you need to have index so you're working here without index so you're getting to your data directly by looking for the Meta data that gives you where you need to get the data.

567
01:20:06.240 --> 01:20:13.950
Josh Moore: And it is a design decision, how much redundancy you build into it right, because all that means you have to keep up you know if you.

568
01:20:14.430 --> 01:20:21.300
Josh Moore: If you delete a well, then you need to go through all the structures and figure out everything that references that well and get rid of it yeah I mean it's.

569
01:20:21.630 --> 01:20:27.270
Jean-Karim Heriche: that's the that's the problem with relational data your you have relationships, you have things that are.

570
01:20:28.020 --> 01:20:44.100
Jean-Karim Heriche: that are linked to each other, they sings points to this image with this Meta data, and you have to keep these together or linked together, so you have to find the balance by between what you keep together and what you what you link.

571
01:20:45.390 --> 01:20:53.910
Jean-Karim Heriche: And, and as I was mentioning at some point somewhere that you will need indices, you will need to index some of the data structure for efficiency.

572
01:20:55.260 --> 01:21:01.620
Jean-Karim Heriche: So maybe you can split some of these things and work out of an index stable or or more than one.

573
01:21:02.670 --> 01:21:07.140
Jean-Karim Heriche: And, and so all these things could be very efficient data structure.

574
01:21:10.980 --> 01:21:16.530
Jean-Karim Heriche: I mean just an idea of following that discussion that's my understanding of it, where it's getting getting.

575
01:21:17.280 --> 01:21:20.220
Josh Moore: And it may be that we can well, there are a couple of ways.

576
01:21:21.570 --> 01:21:30.750
Josh Moore: I assume, initially, we will need to just think about what john crema said, every time we build a spec you know, we had to do it for the hcm the collection speckle have to do it.

577
01:21:31.320 --> 01:21:41.880
Josh Moore: It may be that we can generalize some of this so that there's you know, there is a image index built into Czar or maybe it's a layer higher you know and we build services that.

578
01:21:43.020 --> 01:21:51.330
Josh Moore: Keep the stuff in mind, and if you know if opening your Czar on s3 is just too slow, then you turn on whatever the services that does the indexing.

579
01:21:53.250 --> 01:21:54.510
Josh Moore: kind of remains to be seen.

580
01:21:59.730 --> 01:22:03.480
Josh Moore: anyone else, want to add anything to the collections discussion so.

581
01:22:05.430 --> 01:22:18.240
Kyle Harrington: One thing that i'm trying to understand a bit so part of the issue with collections is at the at the root level, you might have some metadata, it has to apply to the children when you open a child, do you need to have.

582
01:22:18.570 --> 01:22:27.330
Kyle Harrington: be able to get that information and how do you find it so i'm still, after all this not really clear what's the like proposed way of dealing with that.

583
01:22:33.120 --> 01:22:45.660
Josh Moore: So, at the moment, so what at least the hcl specification there you know, there are three levels there's the plate the well and the image level roughly their details, but whatever i'm.

584
01:22:46.890 --> 01:22:52.890
Josh Moore: plate links you to the well well links you to the plate, and the images images links you back to the wells.

585
01:22:56.550 --> 01:22:59.550
Josh Moore: Right, so you so basically bi directional.

586
01:23:00.660 --> 01:23:03.210
Josh Moore: pointers at each of the levels.

587
01:23:05.910 --> 01:23:09.840
Josh Moore: And that's so that all images aren't registered with the plate.

588
01:23:11.190 --> 01:23:19.080
Josh Moore: Right, so if you have to do a two level lookup so it's a collection of collections, which I guess goes back to martin's point of.

589
01:23:19.590 --> 01:23:26.700
Josh Moore: You know, we should we should be able to have collections of anything so in this case we have collections of collections that's the design decision that we made.

590
01:23:27.150 --> 01:23:41.160
Josh Moore: You know, we could roll that back and say no actually a plate should be a full collection and then it would have all the will metadata and all the image metadata all you know all in one go it's that would be valid as well.

591
01:23:49.260 --> 01:24:11.220
Jean-Karim Heriche: I would, I would say I would favor the model where you can ness collections, because the experience shows that you often need these kind of nested structures so flattening might have some advantages, but I think it it loses some of the flexibility, you have when you do I bought one SEC.

592
01:24:18.810 --> 01:24:31.770
William Moore: yeah I guess for the plate we didn't nest too well nesting arbitrarily depth wouldn't make it harder to give an overview of what the what the plate looks like you know you want to be able, quite quickly find.

593
01:24:32.580 --> 01:24:46.980
William Moore: An image to show you for that well and if if the well happened to be have a collection, then you didn't you need to keep iterating a different depth reach so so in at least in that high contents, meaning that that would be harder but yeah it would be more flexible.

594
01:24:54.930 --> 01:24:56.310
Josh Moore: anyone else on collections.

595
01:25:04.290 --> 01:25:05.250
Josh Moore: You have what you need dragon.

596
01:25:08.250 --> 01:25:20.670
Josh Moore: Okay, so knowing we only had about half an hour, and I would like to look at the github repository a bit there any of the other specifications that we didn't touch on that anyone's just dying to.

597
01:25:22.140 --> 01:25:28.950
Josh Moore: Talk about in detail, other than that they want it it's clear all specifications are wanted by someone as quickly as possible so.

598
01:25:30.180 --> 01:25:31.050
Josh Moore: I take that as a given.

599
01:25:35.220 --> 01:25:44.790
Constantin Pape: I just have one procedure, a question so say I want to make a proposal for labels or image spec be to read do I do this now.

600
01:25:45.660 --> 01:25:49.020
Josh Moore: yep that's what I want to end with so definitely won't forget that.

601
01:25:54.180 --> 01:25:59.130
Josh Moore: Shall we just moved to that you guys are always asking the next thing I want to talk about it's great it's just like it just keeps going.

602
01:26:00.600 --> 01:26:04.560
Josh Moore: Okay, I want to make sure no one Okay, so no one's screaming about any other specs.

603
01:26:06.000 --> 01:26:06.690
Josh Moore: Going once.

604
01:26:09.630 --> 01:26:12.330
Josh Moore: awesome okay let's do this, we can always like circle back.

605
01:26:13.710 --> 01:26:21.720
Josh Moore: Oh okay so here's what i've gone through a to a more or less.

606
01:26:23.880 --> 01:26:24.930
Josh Moore: significant degree.

607
01:26:26.130 --> 01:26:27.090
Josh Moore: let's see share.

608
01:26:29.730 --> 01:26:30.300
Josh Moore: So.

609
01:26:34.140 --> 01:26:35.220
Josh Moore: Sorry that's not working.

610
01:26:46.980 --> 01:26:52.590
Josh Moore: Okay, so someone mentioned the the labels specification multi scale image labels one.

611
01:26:53.640 --> 01:26:58.560
Josh Moore: um, and so this kind of kicked off the conversation um.

612
01:27:01.980 --> 01:27:10.530
Josh Moore: And I think only the original multi skilled thread actually got a lot of feedback, so I haven't had a lot of success since then.

613
01:27:11.190 --> 01:27:26.850
Josh Moore: um so both on labels and on hdfs it kind of felt like the Community was going yeah just do what you're doing will will follow suit, which was fine you know that's that's okay um, but it is certainly about driving as much.

614
01:27:28.020 --> 01:27:41.280
Josh Moore: Getting all the disagreements out of the way as early as possible, why does it keep happening um so i'll show you where that did happen and for whoever wants to specs.

615
01:27:42.630 --> 01:27:45.690
Josh Moore: So it's issue 50 So this was the originals our spec work.

616
01:27:47.520 --> 01:27:59.160
Josh Moore: And this went back and forth for a couple of months, you know, and if it works better on github, I guess, I would say, feel free to do it on github as opposed to image sc I try to keep the two.

617
01:27:59.880 --> 01:28:09.300
Josh Moore: apprised of one another, you know but wherever you're managing to get a conversation going do it, and as long as it's on one of those two that's fine um.

618
01:28:10.260 --> 01:28:17.430
Josh Moore: And then, at some point here everyone who participated will remember it started to get into the weeds of.

619
01:28:18.060 --> 01:28:26.460
Josh Moore: How do we represent the Multi necessary, not the Multi scale, but how do we represent the offsets and the position and the grid size and some of the scale information.

620
01:28:27.270 --> 01:28:33.900
Josh Moore: And so, at that point, I just kind of drew drew a line and said okay we're not dealing with that anymore that goes to the next spec.

621
01:28:34.290 --> 01:28:45.960
Josh Moore: So I just want to get that out of the way up front that that's also part of the job is going, you know i'm only going to focus on this much i'm going to help you to you know get to a consensus here and we'll deal with the next bit following that.

622
01:28:47.670 --> 01:28:48.240
Josh Moore: So.

623
01:28:49.590 --> 01:29:03.420
Josh Moore: Once you've kind of driven that conversation, so you know, in this case I would kind of assume you know she posts what she has now in some kind of json representation shows some data sets um.

624
01:29:04.500 --> 01:29:15.990
Josh Moore: And then, wherever that needs to take place that's fine, but I would assume, eventually, it needs to turn into an issue on the N G F F board I assume I hope the dark github.

625
01:29:17.850 --> 01:29:23.190
Josh Moore: style is actually visible over zoom so I find it a bit hard to read um.

626
01:29:25.710 --> 01:29:29.430
Josh Moore: So that would be a pull request against.

627
01:29:30.510 --> 01:29:31.530
Josh Moore: n G F F.

628
01:29:33.150 --> 01:29:34.050
Josh Moore: and

629
01:29:36.840 --> 01:29:43.800
Josh Moore: It would be against latest, so this is currently written in markup called bike shed it's from the w three see.

630
01:29:45.630 --> 01:30:01.740
Josh Moore: The github actions generate the html and everything from this so it's basically just markdown that you can follow on your own um it has some additions for for writing specifications, so you can have a table of definitions, you can have.

631
01:30:03.240 --> 01:30:12.780
Josh Moore: links to other versions of the specifications so anyone who's written a specification, it has you know some of the markup to make it easier, as opposed to meeting do everything in markdown yourself.

632
01:30:14.160 --> 01:30:27.750
Josh Moore: Currently it's all in one file um so we may start conflicting with one another, and if that happens we'll figure out what to do that, you know we'll probably have a multi level specification, which probably gets one what he wants is there's they.

633
01:30:28.410 --> 01:30:30.900
Josh Moore: Of course specifications and then there's.

634
01:30:31.740 --> 01:30:35.400
Josh Moore: You know hts which one can ignore um.

635
01:30:38.700 --> 01:30:41.280
Josh Moore: That PR gets opened.

636
01:30:42.720 --> 01:30:45.720
Josh Moore: And then there should be kind of a period of feedback from the Community.

637
01:30:46.170 --> 01:30:53.190
Josh Moore: i'm MIA culpa I failed to do this, for it okay and sorry this was a conscious decision, and I will do better in the future.

638
01:30:53.550 --> 01:31:03.510
Josh Moore: Basically, the meeting was coming and I wasn't getting feedback on labels and on hdfs, and so we just said okay we're merging this so that's completely known and.

639
01:31:04.230 --> 01:31:19.530
Josh Moore: I will take the bullet for that, so if anyone does have something in those specifications that they hate and it's my fault, and we will fix it in the next version i'm completely valid, but to the extent possible, you know, we really should give everyone time to.

640
01:31:21.120 --> 01:31:38.520
Josh Moore: To have feedback um and to have as many implementations produced for the specs as possible, so that's been the hardest part um so let me go back to the roadmap, and this is just kind of an idea on how we can make this work.

641
01:31:39.810 --> 01:31:49.170
Josh Moore: So this is a github project board for the NGO FF repository um it's called roadmap it's project number one.

642
01:31:50.190 --> 01:31:52.170
Josh Moore: it's linked from the notes, but i'll link it again.

643
01:31:52.740 --> 01:32:10.800
Josh Moore: um so I guess my proposal would be that anything we want to work on so anything that we discussed today, anything that we talked about an image sc or, at any of the Community calls shows up as a card on the roadmap board, this is kind of following if you've not seen it before.

644
01:32:14.640 --> 01:32:28.560
Josh Moore: Get hub project itself has quite a nice roadmap and i've tried to follow that and they just kind of have like all these great colors and things are just like you know, so this is what I see happening right but I need your help, I can't do that myself um.

645
01:32:30.480 --> 01:32:34.020
Josh Moore: You know, put your will put your names on it, you know so.

646
01:32:35.070 --> 01:32:51.870
Josh Moore: I hear coming out of this meeting that we have at least three or four cards that need to live here, they need to have your names on them and um something we haven't talked about a general feeling for when you want to get them merged into the main line right.

647
01:32:53.430 --> 01:33:04.260
Josh Moore: um I don't know how far we want to look, you know, this goes through summer so June of 2021 if someone's planning farther that's fine, we can plan as far as you want in the future.

648
01:33:04.980 --> 01:33:18.870
Josh Moore: um and then the question is what's the bar does it says suffice to have one implementation for a specification, or do we have a set of implementations which must support.

649
01:33:20.070 --> 01:33:38.850
Josh Moore: A new spec before it goes in so concentrate i'm kind of thinking of Czar implementations as a model, you know, so we we take whatever language, whatever implementations we want to support only means our bio formats our reader big data viewer trevor's is rj s whatever.

650
01:33:39.960 --> 01:33:42.810
Josh Moore: And we'll run them through a test suite before we merge.

651
01:33:44.460 --> 01:33:47.400
Josh Moore: thoughts objections extensions.

652
01:33:50.670 --> 01:33:57.360
Constantin Pape: Okay yeah I think that sounds good just sort of where the issue left but it wouldn't be in this on E and F.

653
01:33:58.170 --> 01:33:59.100
Josh Moore: Currently that's where they live.

654
01:33:59.400 --> 01:34:03.150
Josh Moore: You can link elsewhere if that's useful but I, I think, having in one place is the.

655
01:34:03.150 --> 01:34:06.210
Constantin Pape: Best yeah I think that would be better than getting those are.

656
01:34:06.240 --> 01:34:08.550
Constantin Pape: specs involved right, which are, I think.

657
01:34:09.900 --> 01:34:11.700
Josh Moore: yeah that didn't work so.

658
01:34:13.020 --> 01:34:13.860
Josh Moore: I wouldn't do that.

659
01:34:17.880 --> 01:34:24.270
Josh Moore: And everyone so that would mean anyone who's interested can follow this repo should be able to know what's going on.

660
01:34:25.350 --> 01:34:29.670
Josh Moore: I will keep reporting on images see so you know we'll have that bi directional link.

661
01:34:34.410 --> 01:34:39.840
Juan Nunez-Iglesias: Is there, how does one get I guess make the issue to get guidance on.

662
01:34:40.860 --> 01:34:47.550
Juan Nunez-Iglesias: Realistic timelines and consensus because, like you know, to take the scale one as an example.

663
01:34:49.500 --> 01:34:56.070
Juan Nunez-Iglesias: I just don't even I don't even know what you're thinking and and also you know, in terms of the spec.

664
01:34:57.450 --> 01:35:04.530
Juan Nunez-Iglesias: For me, I feel like it's blocking that it's the TC see ya next thing right now and that.

665
01:35:05.790 --> 01:35:09.630
Juan Nunez-Iglesias: The metadata the scale Meta data would have to be somehow tied to that.

666
01:35:11.520 --> 01:35:13.320
Juan Nunez-Iglesias: And I would prefer it wasn't.

667
01:35:14.550 --> 01:35:17.370
Juan Nunez-Iglesias: And so, then yeah so then.

668
01:35:19.200 --> 01:35:21.660
Juan Nunez-Iglesias: Variable number of dimensions, would be an issue that I would make.

669
01:35:22.710 --> 01:35:23.760
Juan Nunez-Iglesias: And then.

670
01:35:24.120 --> 01:35:26.640
Josh Moore: And we do also work on the implementation of the spec change.

671
01:35:28.530 --> 01:35:35.160
Juan Nunez-Iglesias: What does that mean exactly it means adding to the spec and then adding to say armies are pie, because I would that mean.

672
01:35:35.670 --> 01:35:42.450
Josh Moore: Well, I mean to some degree i'm I guess i'm almost less interested in the help on the implementation so that's very welcome, I mean, I think.

673
01:35:43.260 --> 01:35:51.090
Josh Moore: The thing that has to happen first for each of these is really the naming conversation like you have so let's just let's just take this as an example, because the one that.

674
01:35:51.630 --> 01:36:02.940
Josh Moore: The one that i'm one's talking about it's actually fairly simple and it's just a matter of doing it so in multi scale so let's see let's see where does it say five D.

675
01:36:07.890 --> 01:36:11.430
Josh Moore: You know so basically this says.

676
01:36:13.050 --> 01:36:15.840
Josh Moore: You know, it must be five dimensional chunks right.

677
01:36:17.220 --> 01:36:19.380
Josh Moore: So the first question is.

678
01:36:20.520 --> 01:36:28.860
Josh Moore: So this probably so you need to if you're going to change something like this, you need to decide, are you supporting the old version, are you having a breaking version.

679
01:36:29.970 --> 01:36:42.900
Josh Moore: If you want to get your issue done and implemented as quickly as possible don't do it in a breaking style right, I mean that's kind of obvious, so you know I would think what you would do is say.

680
01:36:45.870 --> 01:37:02.490
Josh Moore: You know in you probably do it down below, but you would say something in the text, like if there's not a key dimension order on the array metadata then assume it's five D TC why.

681
01:37:03.570 --> 01:37:04.740
Josh Moore: The wire right.

682
01:37:06.180 --> 01:37:22.770
Josh Moore: If you do find the attribute named whatever dimension order then interpret it this way right, and so, if you can write that up in English that other people understand and everyone's fine with it, then we start looking at the implementations.

683
01:37:29.550 --> 01:37:30.330
Josh Moore: someone's got birds.

684
01:37:43.800 --> 01:37:49.710
Josh Moore: anyone else, seeing how they can fit into this, an issue they want to create and or better implement.

685
01:37:50.670 --> 01:37:59.040
Constantin Pape: I mean i'm so i'm working on an issue for this transformation, I can just share it after so let me take and then everyone can critique.

686
01:38:00.840 --> 01:38:10.440
Constantin Pape: The other thing is, I mean maybe a bit specific about the master this sort of allowing nested is their initial already because I think, for us, which will also be pretty critical.

687
01:38:10.920 --> 01:38:16.020
Josh Moore: So there's not even an issue, yet, for they create one for collections.

688
01:38:17.400 --> 01:38:30.300
Josh Moore: know, so I think the conversation on collections didn't show up as an issue, yet so that already needs to happen, so if someone wants, you know, I was kind of guessing draga would do that so um and then.

689
01:38:32.760 --> 01:38:38.010
Josh Moore: I don't know I assume nested in this can take place within that one issue, maybe we need another I.

690
01:38:38.790 --> 01:38:44.550
Constantin Pape: Think it's a different fun right because it's not about collections, but really about the underlying container.

691
01:38:45.240 --> 01:38:45.480
Maybe.

692
01:38:47.430 --> 01:38:51.090
Kyle Harrington: you're talking about the flat just flat versus nested directories.

693
01:38:51.240 --> 01:38:52.260
Constantin Pape: Exactly yes.

694
01:38:55.950 --> 01:38:56.730
Constantin Pape: For the chance.

695
01:38:58.620 --> 01:38:59.280
Josh Moore: That yes.

696
01:39:00.600 --> 01:39:04.830
Constantin Pape: Yes, and no because that's the part nested storage already.

697
01:39:05.070 --> 01:39:15.930
Josh Moore: It doesn't I know so i've had to open several PR is to get all this to work so zarin jobs are both broken across the board for nested chunks of me.

698
01:39:15.990 --> 01:39:19.830
Constantin Pape: But I mean for me the nested storage yada yada.

699
01:39:21.120 --> 01:39:23.010
Constantin Pape: that's usually so.

700
01:39:23.070 --> 01:39:31.860
Josh Moore: So sorry in five works in five nested on s3 doesn't work geez are doesn't work as our Python fs store doesn't work.

701
01:39:32.850 --> 01:39:35.880
Constantin Pape: And I think there's I pisces one okay.

702
01:39:36.990 --> 01:39:37.890
Constantin Pape: yeah okay.

703
01:39:40.290 --> 01:39:49.290
Josh Moore: But if you, you know if you want to create an issue go for it um and I will link all the stuff I know you're right it desperately needs to happen and.

704
01:39:49.560 --> 01:39:49.800
Constantin Pape: You know.

705
01:39:50.190 --> 01:39:55.590
Josh Moore: And at least that's what i'm fine with doing the implementation yeah so cuz i'm that's where i've been suffering.

706
01:39:56.970 --> 01:40:00.090
Constantin Pape: I guess i'm using my own stuff too much for this.

707
01:40:01.440 --> 01:40:04.680
Josh Moore: yeah if your solution is use the fly for everything that okay.

708
01:40:06.930 --> 01:40:17.550
Eric Perlman: It seems like this has been this is starting to boil over and needs to be addressed by late because I definitely there's been situations where i've used and five instead of Czar based on the pair of tools and languages, I was using.

709
01:40:18.270 --> 01:40:24.180
Eric Perlman: Like I need to Amazon with soulful tools so it's good to be and five on aws.

710
01:40:25.140 --> 01:40:29.280
Josh Moore: yeah that that's kind of the root of all evil at the moment, so if we can fix that that'd be great.

711
01:40:31.710 --> 01:40:50.400
Constantin Pape: Okay sorry this is again a bit unrelated but just because I have these implementations do you think we should maybe much a bring this entity that developers umbrella, because I feel like i'm a bit I mean i'm happy to to work on this, but I think, maybe there should be some.

712
01:40:51.840 --> 01:40:53.730
Constantin Pape: sense of responsibility for it.

713
01:40:53.940 --> 01:40:58.350
Josh Moore: So my you get big thumbs up for me, so I think it's great i'm.

714
01:40:59.010 --> 01:41:08.190
Josh Moore: In general I think this goes for the tsar developers organization and for the oil me organization, you know, to the extent that anyone wants their code to be official.

715
01:41:09.000 --> 01:41:25.980
Josh Moore: that's great you know we love you know you guys are doing great stuff if you want to move it to the org that's fine you probably won't hear anyone say you know, and I thought this about your so can someone paste this our implementations repository please um.

716
01:41:26.130 --> 01:41:29.880
Josh Moore: I fought several times, I should really tell Constantine just to move it to the org.

717
01:41:30.450 --> 01:41:39.090
Josh Moore: But that feels like a very dodgy thing to do, as an org Member, you know you don't feel comfortable going to someone that you know you really should, though.

718
01:41:40.890 --> 01:41:57.450
Josh Moore: I think both of the organizations and, probably, you know, we could say the same thing for Fiji and for all of our party organization if you've developed something that's really cool and you want it to be an official project go for it like you know make that happen that's that's wonderful.

719
01:42:02.910 --> 01:42:05.040
Josh Moore: you're also welcome to just keep it as your own if you know.

720
01:42:06.300 --> 01:42:08.910
Constantin Pape: Making an issue and tagging the relevant people fight.

721
01:42:09.720 --> 01:42:10.110
awesome.

722
01:42:12.990 --> 01:42:25.230
Josh Moore: yeah and I definitely think we need something like that, for me, Czar as well um I don't know if we just copy that repository or if we if anyone has ideas on on how they would want to do that speak up.

723
01:42:28.140 --> 01:42:30.360
Josh Moore: Okay i'm going to look at my list of things that I wanted to say.

724
01:42:30.540 --> 01:42:33.000
Josh Moore: If anyone else has anything they want to add feel free.

725
01:42:34.740 --> 01:42:36.060
Guillaume Gay: Just to make sure.

726
01:42:37.320 --> 01:42:38.820
Guillaume Gay: So far, the mesh if.

727
01:42:40.320 --> 01:42:51.270
Guillaume Gay: The correct workflow is, first, the specification, then the implementation so I should have been an issue to specify how to store meshes here.

728
01:42:52.440 --> 01:42:53.160
Josh Moore: So.

729
01:42:55.470 --> 01:43:03.180
Josh Moore: Help them recover couple of questions built into that so completely feel free to open an issue that would be wonderful to so big support um.

730
01:43:03.780 --> 01:43:15.840
Josh Moore: To be completely honest, you know I we probably don't always do this specification, first we often start with you know the that at least one implementation first just because you know we're human um but you know it.

731
01:43:17.730 --> 01:43:29.430
Josh Moore: A specification should come with one implementation and generally maybe maybe i'm wrong i've found that by pasting a very simple snippet of json.

732
01:43:30.510 --> 01:43:36.000
Josh Moore: people understand what i'm talking about better, you know that that seems so a screenshot saying this is.

733
01:43:36.000 --> 01:43:36.600
Guillaume Gay: cool.

734
01:43:36.780 --> 01:43:46.290
Josh Moore: And here's the way I want to name an end json gets gets a lot of people involved so that's what I would suggest, but feel free to play around with it i'm.

735
01:43:46.320 --> 01:43:46.680
Josh Moore: You know.

736
01:43:47.520 --> 01:43:48.450
Josh Moore: we're all new to this.

737
01:43:53.010 --> 01:44:13.170
Jean-Karim Heriche: But I, I agree with josh I think the specification should come with a reference implementation because of a specification always as ambiguities and the reference implementation basically takes care of that it, it shows what actually is meant in when it's a big years.

738
01:44:14.190 --> 01:44:22.380
Jean-Karim Heriche: And it can be used for testing all the implementations making sure that they do the same and the top with the specification, the same.

739
01:44:26.220 --> 01:44:26.490
Guillaume Gay: Okay.

740
01:44:34.860 --> 01:44:36.090
Josh Moore: anyone else have anything they want to.

741
01:44:37.740 --> 01:44:38.340
Josh Moore: throw out there.

742
01:44:43.980 --> 01:44:45.600
Josh Moore: shouldn't be that mean to look critical.

743
01:44:52.830 --> 01:45:03.390
Josh Moore: So I think the last question from my point is whether or not anyone feels that needs to be anything different or more frequent or whatnot in terms of communication.

744
01:45:04.350 --> 01:45:22.410
Josh Moore: um I think the standing agreement is that every you know roughly once a month will have one of these calls i've kind of failed and it's been more like every two months, which actually has felt okay as well, but if I plan on every month and I actually make every two months happen um.

745
01:45:24.120 --> 01:45:40.860
Josh Moore: It may be that if more people are building the specs will need to talk to the people who are actively working on trying to merge into the main line will need to chat more often I would leave it to each of you to just raise a flag, when you want to chat and we'll invite whoever.

746
01:45:42.480 --> 01:45:43.290
Josh Moore: um.

747
01:45:44.340 --> 01:45:46.800
Josh Moore: Can formalize that if we want, but I figured that's yeah.

748
01:45:51.030 --> 01:46:03.450
Jean-Karim Heriche: Do you have a chat channel some something for for people will will actually get their hands dirty on the on the specifications on the implementations where they actually can.

749
01:46:04.650 --> 01:46:12.660
Jean-Karim Heriche: Directly get in touch with each other, without necessarily going VI issues or us who image see.

750
01:46:13.260 --> 01:46:24.360
Josh Moore: I don't know so good question very good question and open to opinions, the two things I know of, and if anyone wants to suggest more ones already smiling i'm.

751
01:46:25.530 --> 01:46:31.200
Josh Moore: The first would be github discussions I don't know if anyone has had good experience with those um.

752
01:46:32.430 --> 01:46:37.950
Josh Moore: Certainly, for Czar it doesn't tend to work, you know, so I don't get much feedback when I use the Czar discussions.

753
01:46:38.670 --> 01:46:49.080
Josh Moore: But if everyone is watching a basically a github discussion looks like an issue but there's no assumption that it's an issue it's just it's a place where you can chat so.

754
01:46:49.650 --> 01:46:57.660
Josh Moore: that's one option, the other option which i've proposed a couple of times so one time, I proposed it formally, which was a separate julep.

755
01:46:58.680 --> 01:47:10.950
Josh Moore: Beside the image se discourse, because sometimes you just need chat um there wasn't a general acceptance of that from the administrators of the discourse, which is fine.

756
01:47:12.030 --> 01:47:26.280
Josh Moore: Since various rogue people like one have suggested it again and again, and so there is an image se julep I don't mind using it um but it kind of yeah that requires a larger decision.

757
01:47:28.020 --> 01:47:29.160
Juan Nunez-Iglesias: On there if people want to.

758
01:47:30.990 --> 01:47:31.470
Josh Moore: Like I said.

759
01:47:33.600 --> 01:47:34.560
Josh Moore: yeah well yeah.

760
01:47:36.090 --> 01:47:37.230
Josh Moore: i'm not guilt free it's.

761
01:47:42.690 --> 01:47:45.240
Josh Moore: Well, if github discussions work I think they're.

762
01:47:45.240 --> 01:47:46.020
Constantin Pape: Not sorry I.

763
01:47:46.380 --> 01:47:46.830
Juan Nunez-Iglesias: Get her.

764
01:47:46.860 --> 01:47:47.190
Juan Nunez-Iglesias: Get her.

765
01:47:47.250 --> 01:47:47.730
Constantin Pape: Get her.

766
01:47:47.940 --> 01:48:00.810
Josh Moore: I get her, I think the threading on guitar is still very painful I can I can never find or follow the threads on guitar and the one thing that zoom does really well.

767
01:48:01.350 --> 01:48:10.950
Josh Moore: much better than slack or anything else is named conversations you know so every thread is clearly named and you know what you're talking about and I love that.

768
01:48:11.010 --> 01:48:11.580
Juan Nunez-Iglesias: You know so.

769
01:48:12.000 --> 01:48:15.060
Josh Moore: For example, when it comes to the party zoo look like I only.

770
01:48:15.420 --> 01:48:18.540
Josh Moore: Read the ones that have all me and the title and I ignore everything else.

771
01:48:23.490 --> 01:48:24.660
Kyle Harrington: Whereas like worse than.

772
01:48:31.980 --> 01:48:32.940
Josh Moore: Someone faster than me.

773
01:48:33.960 --> 01:48:34.260
Josh Moore: Oh.

774
01:48:39.780 --> 01:48:42.360
Josh Moore: would have been fine my screensaver came on anyway.

775
01:48:47.610 --> 01:48:52.800
Josh Moore: So we can try it out, we can veto that though so that's that's the that's the.

776
01:48:54.360 --> 01:48:55.740
Josh Moore: pirate conversations.

777
01:48:59.490 --> 01:49:02.220
Josh Moore: Basically i'll respond to any CC anywhere so just.

778
01:49:03.000 --> 01:49:12.450
Juan Nunez-Iglesias: It does it does have premier links as well with the caveat that you do need an account, and you have to you can sign in with github or whatever, but it does.

779
01:49:19.800 --> 01:49:30.780
Josh Moore: anyone else have a preference or a suggestion for a place to do this, so my first example of a project, leaving get her to go to discord this morning I was like Oh, my goodness.

780
01:49:33.120 --> 01:49:33.720
Juan Nunez-Iglesias: I will.

781
01:49:34.740 --> 01:49:44.010
Juan Nunez-Iglesias: send me advertise the thing that i'm working on, but hasn't caught off the ground, yet which is in Jesse island modeled after I tick island.

782
01:49:45.210 --> 01:49:45.750
Josh Moore: So this is.

783
01:49:47.700 --> 01:49:48.540
Josh Moore: gather town.

784
01:49:48.990 --> 01:49:49.290
yeah.

785
01:49:55.980 --> 01:50:13.530
Deborah Schmidt: I would personally support using the phone for most cases, because you can just read up on it and sync asynchronously and I, for example, like we I told me over to get the last month and I just cannot keep up with all the communication channels at all, it just this so much.

786
01:50:14.730 --> 01:50:17.490
Josh Moore: So that's actually a good point Deborah all right.

787
01:50:18.510 --> 01:50:27.840
Josh Moore: i've struggled with this What could we do an image sc like do we need a new category or a new tag to make people feel comfortable to just.

788
01:50:29.040 --> 01:50:37.860
Josh Moore: Use discourse like it's a chat because I discourse is great, like, I think it works wonderfully but i've had a hard time getting people to feel comfortable just.

789
01:50:37.890 --> 01:50:38.670
Deborah Schmidt: yeah I agree.

790
01:50:39.240 --> 01:50:43.470
Josh Moore: You know, so if someone knows how to fix that barrier i'm all for staying on discourse.

791
01:50:43.950 --> 01:50:49.980
Deborah Schmidt: I thought about that for a while and I think it would be nice to have a better starting page that you could personalize.

792
01:50:50.610 --> 01:51:05.250
Deborah Schmidt: And where you could say, these are the texts that i'm interested in and like kinetic organize them and then maybe see where take then what's unanswered and things like that and tidy looked a bit into it and you can write plugins for this course, but I think we are using.

793
01:51:06.570 --> 01:51:09.750
Deborah Schmidt: I like we don't say first right.

794
01:51:10.860 --> 01:51:18.420
Deborah Schmidt: it's like subliminal paid version, and they are limited, when it comes to writing your own plugins are completely restricted.

795
01:51:18.930 --> 01:51:21.000
Deborah Schmidt: So I think that can show what we have.

796
01:51:21.360 --> 01:51:32.160
Josh Moore: We pay quite a lot and they're very responsive, so if we say we want something there's a developer from discourse who sits in our slot our discourse and response so.

797
01:51:33.240 --> 01:51:35.640
Josh Moore: You know if you yeah so if.

798
01:51:36.960 --> 01:51:41.370
Josh Moore: it's a bit weird to have it be all commercial but on the other hand, it just kind of work so.

799
01:51:41.910 --> 01:51:48.240
Deborah Schmidt: yeah I totally understand that I just from what I read, I read that it's limited when.

800
01:51:49.320 --> 01:51:52.620
Deborah Schmidt: that's not the case that's that could really be an opportunity to.

801
01:51:53.670 --> 01:51:57.480
Deborah Schmidt: make it more person and less and treating like let's.

802
01:51:58.590 --> 01:52:03.450
Deborah Schmidt: Say that people go there and you just have a list of huge topics and then they leave again right.

803
01:52:03.900 --> 01:52:13.710
Josh Moore: So I don't know how, so this is why I created the in gfs group because basically people were saying i'm not seeing what you're saying josh I don't know what's going on, I was like crap you know so it's like.

804
01:52:14.040 --> 01:52:20.370
Josh Moore: We have we've brought our whole community multiple communities together and now, people are lost, and like okay that's bad.

805
01:52:20.820 --> 01:52:21.840
Josh Moore: So, if anyone knows about.

806
01:52:23.280 --> 01:52:24.000
Josh Moore: So if you know.

807
01:52:24.660 --> 01:52:28.770
Deborah Schmidt: yeah I really would like to to to contribute their.

808
01:52:31.980 --> 01:52:41.430
Josh Moore: Thank you kyle So yes, like in send it my way and we'll try to get it installed and then just try it out awesome for support.

809
01:52:43.890 --> 01:52:45.570
Josh Moore: Okay, four minutes last call.

810
01:52:48.090 --> 01:52:48.990
Josh Moore: You still have your whiskey.

811
01:52:49.020 --> 01:52:49.410
one.

812
01:52:52.110 --> 01:52:54.540
Juan Nunez-Iglesias: it's all gone I do need to go down for another.

813
01:52:58.560 --> 01:52:59.520
Josh Moore: was good, seeing everyone.

814
01:52:59.760 --> 01:53:00.510
Josh Moore: Thank you for coming.

815
01:53:01.830 --> 01:53:02.220
To keep.

816
01:53:03.630 --> 01:53:05.220
Josh Moore: you'll see the notes on images see.

817
01:53:06.690 --> 01:53:07.950
Josh Moore: don't be a stranger.

818
01:53:08.310 --> 01:53:08.880
Guillaume Gay: cetera et cetera.

819
01:53:11.100 --> 01:53:12.180
Juan Nunez-Iglesias: So thanks for organizing.

820
01:53:12.990 --> 01:53:14.700
Josh Moore: i'll stick around if anyone wants to say.

821
01:53:19.290 --> 01:53:19.950
Jason Swedlow: Good see well.

822
01:53:22.560 --> 01:53:22.920
Eric Perlman: I was.

823
01:53:23.250 --> 01:53:25.890
Eric Perlman: I was curious how's your durable answer tinkering going.

824
01:53:26.700 --> 01:53:28.020
Josh Moore: Well, I can get it running and all.

825
01:53:29.880 --> 01:53:40.230
Josh Moore: But I can't figure out user interface, so I can get my Czar loaded but it doesn't like the five dimensions are I assume it's wanting a three dimensional is our.

826
01:53:40.350 --> 01:53:55.050
Eric Perlman: room it'll take n dimensional if you click I think I left the link at some point on them it's forum, but the key thing is the Jason I use their for mapping dimension, so it wants it'll take n dimensional, but it really wants to know what X, Y Z and car.

827
01:53:55.320 --> 01:54:02.400
Josh Moore: So I tried my best to teach it what X, Y Z and Z was but it didn't learn So if you could send me the json that would be lovely.

828
01:54:02.790 --> 01:54:05.940
Eric Perlman: Okay yeah the interface is not at all intuitive.

829
01:54:05.970 --> 01:54:07.500
Eric Perlman: It is it's.

830
01:54:07.710 --> 01:54:14.700
Eric Perlman: Totally functional but it's a translation matrix of this dimension maps to this dimension and have you noticed it's totally bizarre.

831
01:54:15.510 --> 01:54:19.740
Josh Moore: But yeah it should it shouldn't so I went for the inverse identity matrix and it.

832
01:54:19.740 --> 01:54:20.370
Josh Moore: didn't work.

833
01:54:20.880 --> 01:54:24.390
Eric Perlman: Well, I ended up needing to do not even the inverse identity matrix they needed.

834
01:54:25.770 --> 01:54:31.350
Eric Perlman: Well okay depends on which state I don't think I what I did for bio formats to rile put, which is what you're probably doing this on.

835
01:54:32.700 --> 01:54:34.590
Eric Perlman: The main one is channel.

836
01:54:36.000 --> 01:54:42.900
Eric Perlman: What you can't do see carrot unless all the channels and trump together right now I did finally create a neuro plantation for that.

837
01:54:43.200 --> 01:54:52.200
Eric Perlman: Because JEREMY JEREMY he's the only one of the universe will ever be able to implement that whereas I would be able to figure out how to at least read the old me tiff or omens our metadata and do something with it.

838
01:54:54.420 --> 01:54:59.550
Josh Moore: find the link for it at this little bit ago yeah so 284 so i'll follow that she went.

839
01:55:03.660 --> 01:55:06.000
Eric Perlman: To see your own posts on the image se forum.

840
01:55:07.440 --> 01:55:09.750
Josh Moore: He certainly with search.

841
01:55:12.600 --> 01:55:16.080
Josh Moore: The search is really good and discourse if you learn all the.

842
01:55:17.190 --> 01:55:24.300
Josh Moore: There are a ton of keywords so i've i've added a couple PRS to discourse it's pretty well done, oh.

843
01:55:31.770 --> 01:55:36.240
Eric Perlman: yeah so i'm going to paste this thread.

844
01:55:38.220 --> 01:55:45.960
Eric Perlman: I had a little link a link in there that says drug lancer DEMO if you click that and then click on the top right corner the braces you can see.

845
01:55:48.780 --> 01:55:55.980
Eric Perlman: You can look at the Jason and, among other things, you could copy and paste roughly where I described the layers or you know.

846
01:55:58.110 --> 01:56:00.180
Eric Perlman: The dimension i'm not sure which part of this.

847
01:56:00.510 --> 01:56:04.200
Josh Moore: dimension because that's empty so layers.

848
01:56:06.150 --> 01:56:10.710
Josh Moore: Is no fillers fillers transform meet.

849
01:56:13.020 --> 01:56:16.650
Eric Perlman: You should be able to have more or less copy and paste that directly in the middle three parts it.

850
01:56:20.700 --> 01:56:31.980
Eric Perlman: it's one thing to be clear about nerd glance or no one far as I can tell, except JEREMY actually ever uses the interface for doing any of this all like if the Sebastian sons lab and everyone else doing things based on neuroscience or.

851
01:56:32.130 --> 01:56:36.510
Eric Perlman: In almost all in Python up to doing Jason and updating the States.

852
01:56:37.680 --> 01:56:39.630
Eric Perlman: Is the main interaction, people are doing with it.

853
01:56:42.480 --> 01:56:44.430
Josh Moore: good to know okay i'll give it to play, thank you.

854
01:56:45.300 --> 01:56:48.690
Eric Perlman: I like the model of the way of dealing with the dimensions and whatnot.

855
01:56:49.800 --> 01:56:57.390
Eric Perlman: yeah do you want most of you have multiple channels and the same chunk it has a cute little feature where you can name it see what the carrot.

856
01:56:57.660 --> 01:57:06.360
Eric Perlman: And then the shaders get that as dimensions zero the end so you can have a shader that combines channel 01 and two and rgb, for example in the word three channel image.

857
01:57:07.560 --> 01:57:08.040
Eric Perlman: I noticed.

858
01:57:09.330 --> 01:57:13.230
Eric Perlman: A lot of activity of the bio formats to raw in the last like week.

859
01:57:13.560 --> 01:57:20.010
Eric Perlman: i'm excited about chug sizes, that I was very happy to see you're mentioning that chunk sizes should extend the things other than X and y.

860
01:57:20.340 --> 01:57:25.380
Josh Moore: yeah I think I mean I think we're well we're focusing on dead so.

861
01:57:26.580 --> 01:57:27.180
Josh Moore: i'm.

862
01:57:28.650 --> 01:57:29.400
Josh Moore: David wood.

863
01:57:29.940 --> 01:57:32.520
Eric Perlman: Right, but what you're generalizing for Z like.

864
01:57:34.350 --> 01:57:39.570
Josh Moore: Well, except it's a bit worse so Z is worse because you have to do the down sampling.

865
01:57:41.550 --> 01:57:44.970
Josh Moore: Or we want to do the down sampling and then and then.

866
01:57:46.020 --> 01:57:51.330
Josh Moore: Changing the chunk size of see should just be kind of are you know super simple.

867
01:57:52.440 --> 01:57:54.870
Josh Moore: I think, David can scream at me if i'm wrong.

868
01:57:59.100 --> 01:57:59.880
David Gault: yeah I mean.

869
01:58:02.190 --> 01:58:12.150
David Gault: The issue with doing the done like chunk size and see is more legacy file formats issues, the elections are issues at the minute.

870
01:58:13.530 --> 01:58:17.460
David Gault: You know, actually implementing started shopping in multiple dimensions isn't.

871
01:58:19.020 --> 01:58:20.640
David Gault: isn't too big, of an issue for.

872
01:58:21.660 --> 01:58:27.750
Eric Perlman: isn't channel basically the outer loop of almost any of the integrators or is it, I guess, time and channel.

873
01:58:30.300 --> 01:58:34.650
Eric Perlman: I guess I didn't I didn't work on data sets with one of the one time went usually so I guess I kind of ignore that.

874
01:58:37.410 --> 01:58:43.110
Eric Perlman: The down sampling, I guess, I didn't could have brought it up you're mentioning multi scales earlier that is it's still like.

875
01:58:43.680 --> 01:58:52.950
Eric Perlman: not clear whether one always one wants to be able to both down sample in just XY plane, or sometimes in xyz planes and so being able to.

876
01:58:53.550 --> 01:59:04.350
Eric Perlman: kind of easily do one or the other sort of matters and the em world generally is down sample in X and y until you get ice a Tropic and then you do all three.

877
01:59:06.390 --> 01:59:07.170
Josh Moore: Interesting okay.

878
01:59:08.580 --> 01:59:15.780
Sebastien Besson: does depend on your volume essentially if you got something which is largely ISO Tropic generally you're doing X, Y Z.

879
01:59:17.490 --> 01:59:33.540
Sebastien Besson: You know, it depends on how Fini essentially how thin your initial structure is essential, even the the whole doubt, I mean wsi word good you're largely to the can be stuck but really you're more to the event anything so while the lead sheet and the emmys here's.

880
01:59:33.600 --> 01:59:36.330
Sebastien Besson: My big block So if I start only doing ex wife and.

881
01:59:36.780 --> 01:59:39.540
Sebastien Besson: I actually don't benefit so much from it down sampling.

882
01:59:41.730 --> 01:59:57.360
Eric Perlman: yeah I think with so that's what said they'll dvm the suit that sorry cereal section em is usually the first couple layers are just X, Y until you get it Tropic ish and then it's maintained, whereas like get them some the whole way you would do it.

883
01:59:58.920 --> 02:00:02.040
Eric Perlman: Anyway, I want to go to bed it's late thanks for having us.

884
02:00:02.280 --> 02:00:02.580
Josh Moore: well.

885
02:00:02.640 --> 02:00:04.920
Sebastien Besson: Are you coming later i'll you see it coming tonight.

886
02:00:05.580 --> 02:00:07.830
Eric Perlman: No, no, I will not, I actually have like.

887
02:00:07.890 --> 02:00:13.590
Eric Perlman: This is the first day of which i've now been have been apple aligned meetings at my schedule at that time.

888
02:00:14.340 --> 02:00:27.630
Eric Perlman: Which is going to be fun to navigate because there's two of them that i'm supposed to absolutely attend so i'm going to do the zoom on one window webex, on the other, and see our associates, and unfortunately I don't think I can add, and you have that on my phone, for good measure.

889
02:00:30.180 --> 02:00:31.710
Josh Moore: You get bonus points, if you show up.

890
02:00:33.780 --> 02:00:35.220
Eric Perlman: My cat may show up how's that.

891
02:00:35.250 --> 02:00:36.570
Josh Moore: Okay that's me, I will take a character.

892
02:00:36.960 --> 02:00:37.410
Eric Perlman: There yeah.

893
02:00:39.840 --> 02:00:40.980
Eric Perlman: Okay i'm good I guys.

894
02:00:41.460 --> 02:00:41.850
tonight.

895
02:00:45.630 --> 02:00:46.050
Juan Nunez-Iglesias: as well.

896
02:00:46.350 --> 02:00:46.920
Josh Moore: All right, take care.

897
02:00:47.130 --> 02:00:47.670
Everyone.

898
02:00:50.280 --> 02:00:50.790
Nicolas Vinuesa: Thank you.

899
02:01:04.140 --> 02:01:04.410
Josh Moore: All right.

900
02:01:05.970 --> 02:01:06.630
Josh Moore: You guys take care.