WEBVTT

1
00:00:00.000 --> 00:00:03.899
Josh Moore: Alright, I'm kicking off recording, actually. Hopefully there's enough space on the cloud.

2
00:00:05.790 --> 00:00:16.980
Josh Moore: So now that recording is running. I'll just say that everyone who's here agreed that we could record them so they complain. Later, we now have proof. Um, okay. So a few kind of

3
00:00:18.480 --> 00:00:25.860
Josh Moore: Getting Started comments. Has everyone found access to the hacking D document.

4
00:00:29.430 --> 00:00:38.610
Josh Moore: Paste it into the chat, it's this beautiful thing. Hopefully you've played around with hacking MD before it's really nice to take all of our notes in

5
00:00:39.330 --> 00:00:45.120
Josh Moore: markdown and be able to publish them pretty instantly. So that's why we're doing this, I'm

6
00:00:45.660 --> 00:00:59.580
Josh Moore: To the extent possible, try to add stuff to the document like if you have questions or you want to add topics, anything you want us to try to cover in the next two hours if you can get it there as opposed in the zoom chat room. That'll make things simpler, but if need be.

7
00:01:01.410 --> 00:01:04.110
Josh Moore: Put something in chat and then someone else will copy it over.

8
00:01:06.240 --> 00:01:06.840
See

9
00:01:11.970 --> 00:01:13.950
Josh Moore: I actually don't know who Igor is

10
00:01:17.130 --> 00:01:18.030
Josh Moore: Can someone check the

11
00:01:26.460 --> 00:01:26.760
Jason Swedlow: Intro

12
00:01:28.350 --> 00:01:33.810
Josh Moore: Thank you. Yeah. Someone figure out if we know who Igor is and then let them in tears. Um,

13
00:01:35.100 --> 00:01:41.730
Josh Moore: Alright, so some of you have started adding information about what you're working on, um,

14
00:01:42.570 --> 00:01:51.900
Josh Moore: What you need help with. So there are blocks higher up in the document to cover those partially. That's what we're trying to do here is get everyone together. Find out how as a community, we can move towards

15
00:01:53.580 --> 00:01:58.890
Josh Moore: This concept of next generation data access, you know, it doesn't necessarily have to be file formats.

16
00:01:59.460 --> 00:02:07.050
Josh Moore: Um, so it'd be good to get an overview from as many people as possible. So we'll go around to quick introductions do try to keep it to a minute.

17
00:02:07.410 --> 00:02:17.310
Josh Moore: So you can start adding notes about yourself in any of the locations in the document, you know, if you have details you have links that you want us to look at and by all means, add them.

18
00:02:17.880 --> 00:02:33.240
Josh Moore: And everyone can you know asynchronously go through them and look, while the rest of the chatting is going on arm and then we'll start diving into high level questions or anything we want to talk about more and figure out how we're going to route run this kind of

19
00:02:34.440 --> 00:02:35.970
Josh Moore: This experiment. So

20
00:02:39.150 --> 00:02:52.800
Josh Moore: Yeah, it's the first of the Community calls. Welcome to the inaugural one, I guess you all get to help shape how we're going to do these and to what extent we keep doing them. All right, I will start going through everyone in the order on my participants list.

21
00:02:55.140 --> 00:03:00.300
Will Moore: Assuming everyone's looking at their own copy of the hack ND, you probably need to share your screen.

22
00:03:01.080 --> 00:03:02.910
Josh Moore: And share. Thanks. Well,

23
00:03:05.820 --> 00:03:06.540
Josh Moore: All right, I'm gonna

24
00:03:08.760 --> 00:03:16.500
Josh Moore: Let all of our kind not non oil me guests on dundonians go first and skip the hosts.

25
00:03:17.880 --> 00:03:19.890
Josh Moore: And so that leaves it alphabetical with Anatol

26
00:03:23.790 --> 00:03:24.120
First

27
00:03:25.530 --> 00:03:25.920
Josh Moore: Yes.

28
00:03:28.290 --> 00:03:34.440
Anatole Chessel: Hi. So yeah, news, I thought to sell I'm assistant professor at the ecliptic nick, nick Paris.

29
00:03:36.570 --> 00:03:43.200
Anatole Chessel: I work on well but even imagined fanatics. I had the pleasure of interacting with

30
00:03:44.310 --> 00:03:46.080
Anatole Chessel: People when I was a postdoc as

31
00:03:47.400 --> 00:04:02.640
Anatole Chessel: I worked with the audio portion of it bags and I worked on a Hacker's cleaning our work more on large data set. And I'm interested in this code for tourism, when is that the line by line. I mean, is also finally

32
00:04:03.690 --> 00:04:20.610
Anatole Chessel: Revealed you Moscow about our fancy book face which button particular emerging kind of a scope which is feel proud of that which we need to start somewhere. And for now, we don't know how to that comes all the way to it, I think. So it's much

33
00:04:20.640 --> 00:04:21.240
Will Moore: More later.

34
00:04:21.570 --> 00:04:35.670
Anatole Chessel: Image producer. What sort of metadata, we want. And how do you want to store things. And the second reason is that we will soon publish new person every which a matter

35
00:04:37.290 --> 00:04:42.270
Anatole Chessel: Helping at doing quantitative metric. Notice button.

36
00:04:44.580 --> 00:04:46.380
Josh Moore: Is hitting you with a stick and at all.

37
00:04:46.440 --> 00:04:48.270
Sebastien Besson: I'm hitting you. The one minute stick

38
00:04:49.290 --> 00:05:00.570
Anatole Chessel: And so, which which was I created my own interest is how do we store to the data on the wall so meshes and the trees and the curves and

39
00:05:02.100 --> 00:05:05.820
Josh Moore: Awesome, thank you tried to go first. All right, Christian

40
00:05:07.050 --> 00:05:08.250
Josh Moore: Get a top end up on time.

41
00:05:09.210 --> 00:05:22.620
Christian Tischer: So I'm a Christian. I work at em Heidelberg, and we have a lot of big image data and we are also sharing this remotely remotely and in the cloud, but we are using a

42
00:05:23.730 --> 00:05:35.970
Christian Tischer: And five data. And I think I would rather use to something like to use something which I would all the communities to nerd, which I hope we can establish so that my dream for the next five years.

43
00:05:37.410 --> 00:05:46.530
Christian Tischer: And my motivation is daily suffer of all these fun robots. So if, if this would go out of my life would make me very happy.

44
00:05:48.810 --> 00:05:51.900
Josh Moore: So yeah, that's it. Thanks, Dave.

45
00:05:55.590 --> 00:06:02.340
Dave Mellert: Hi, my name is Dave Miller and I usually look a little more put together but I slept through my alarm. So here I am.

46
00:06:04.380 --> 00:06:12.090
Dave Mellert: I work in research it at the Jackson Laboratory and I lead a small group that's very researcher facing we work with

47
00:06:13.410 --> 00:06:15.300
Dave Mellert: The scientific community to

48
00:06:16.320 --> 00:06:29.910
Dave Mellert: help them identify you know workflows and tools that can help them solve their image analysis problems and and then we help get those things deployed in our environment we train them on their use.

49
00:06:31.650 --> 00:06:33.300
Dave Mellert: We have a lot of folks that want to do

50
00:06:34.620 --> 00:06:44.010
Dave Mellert: High Throughput and highly paralyzed image processing and our HTC environment and in the cloud. And that's why I'm very interested in this topic.

51
00:06:48.300 --> 00:06:58.440
David Pinto: Hi. So my name is David. I work at micron Oxford and I'm working on two projects called microscope and cockpit and basically their

52
00:06:59.520 --> 00:07:03.990
David Pinto: Software for controlling microscopes, with a focus on arbitrary complex

53
00:07:04.440 --> 00:07:12.870
David Pinto: And very exotic devices and idea that, you know, biologists will be able to just write your experiments because they just become Python scripts, so I'm

54
00:07:13.290 --> 00:07:26.550
David Pinto: Interested on these because they're not appointed like okay, that seems starting to work. And now we need to store the data in some format and you know if this is next generation. We don't want to start with the old one to in two years time change it so

55
00:07:27.630 --> 00:07:28.440
David Pinto: He goes next

56
00:07:30.240 --> 00:07:41.100
Egor Panfilov: Hello everyone, my name is Igor. I'm doing my PhD at University of all in Finland, I have background image processing computer vision in industry and now

57
00:07:42.210 --> 00:07:48.450
Egor Panfilov: I'm working with medical images, mainly. I'm also part of the core chemo psychic image so

58
00:07:50.400 --> 00:07:56.880
Egor Panfilov: I joined this image in the short notice. So at this point, my main interest is to learn more about czar and see how

59
00:07:57.990 --> 00:08:03.690
Egor Panfilov: How we can use it in in our daily work and all since I can. So thank you for joining.

60
00:08:05.430 --> 00:08:05.670
Egor Panfilov: Me.

61
00:08:06.300 --> 00:08:15.810
Emil Rozbicki: My name is Elizabeth camera head of applications for Glencoe software arm so my main interest is to learn a little bit more about you guys obviously in your clothes and probably share our experiences.

62
00:08:16.410 --> 00:08:31.200
Emil Rozbicki: Especially from the commercial side or we see all the big companies moving into cloud as well. And the hybrid solutions, especially for high content screening call slide images multiple multiplex damages so so very demanding space. Thanks, guys.

63
00:08:32.550 --> 00:08:38.760
Eric Perlman: My name is Eric by introduction to this world was I didn't really have where I work on the project to collect the full library of

64
00:08:39.000 --> 00:08:52.860
Eric Perlman: TM my introduction to end five and Tsar was Stefan Sommerfeld converting the entire brain two and five format so he could have rapid access to volumetric regions at whole resolution levels, which proved to be very useful.

65
00:08:53.490 --> 00:09:05.970
Eric Perlman: Currently, I've been contracting with Jax with Dave Miller here and working on access to data stored in America. So it's been an introduction to America, but it's been fun to see that you're going to these two worlds.

66
00:09:08.940 --> 00:09:18.660
Erin Diel: Hi everyone. I'm Erin. I'm an application specialist with Glencoe I work really closely with a meal. So I'll save your time in just second but he shared and again here to learn from everyone, and what you're all working on

67
00:09:20.250 --> 00:09:20.580
Erin Diel: Things.

68
00:09:21.090 --> 00:09:23.460
Josh Moore: You are too fast, that didn't come up in the next person to

69
00:09:26.280 --> 00:09:31.200
Guillaume Gay: Hire and young so I'm like a bit new tool that

70
00:09:32.430 --> 00:09:38.730
Guillaume Gay: I'm charged with setting up a zero for the whole Rimini campus in now say

71
00:09:39.810 --> 00:09:52.860
Guillaume Gay: So lots of different labs subjects in and tools, custom, custom microscopes custom night shift microscopes and

72
00:09:54.270 --> 00:09:57.210
Guillaume Gay: Lots of neuro scientists who have a whole

73
00:09:58.500 --> 00:09:59.190
Guillaume Gay: Different

74
00:10:01.140 --> 00:10:06.000
Guillaume Gay: zoology of data. So I'm here to learn about czar and see how I can

75
00:10:07.020 --> 00:10:10.560
Guillaume Gay: Use that for all those very different questions.

76
00:10:12.720 --> 00:10:13.050
Thank you.

77
00:10:16.680 --> 00:10:16.860
Josh Moore: Oh,

78
00:10:18.180 --> 00:10:18.450
Jean-Karim Heriche: Yeah.

79
00:10:19.200 --> 00:10:34.920
Jean-Karim Heriche: Hi I'm Michelle I am I work at end in Hyderabad and I work for the cell biology and biophysics unit I typically provide computational support in various areas, but mostly dealing with

80
00:10:36.180 --> 00:10:50.880
Jean-Karim Heriche: Data downstream from image enterprises, I call it cleaning up the mess but also lately I've been involved in the European side Open Science cloud project, and in particular I'm work package leader in

81
00:10:51.990 --> 00:10:53.340
Jean-Karim Heriche: IOS life which

82
00:10:55.110 --> 00:11:01.770
Jean-Karim Heriche: Tries to implement the European Open Science cloud for the life sciences and here I'm mostly interested in

83
00:11:03.600 --> 00:11:10.260
Jean-Karim Heriche: How to do image analyzes and do data analysis in the cloud. So to deploy.

84
00:11:11.370 --> 00:11:16.170
Jean-Karim Heriche: Imaging software they imaging data in the cloud and how to deal with them.

85
00:11:21.240 --> 00:11:21.990
Josh Moore: Leaving one

86
00:11:23.130 --> 00:11:23.580
Juan Nunez-Iglesias: Hello.

87
00:11:24.780 --> 00:11:25.680
Juan Nunez-Iglesias: So I'm on

88
00:11:27.330 --> 00:11:36.540
Juan Nunez-Iglesias: I'm at Monash University in Australia. I work on second image in the party and a few other things, and a few months ago.

89
00:11:37.260 --> 00:11:49.770
Juan Nunez-Iglesias: The Army guys did the army's our plugin for the party, and it's just makes working with large images so easy and so I'm here to see how I can improve that format so I can do things like

90
00:11:51.090 --> 00:11:56.880
Juan Nunez-Iglesias: Add scale to it and yeah generally make it be the only format that I have to deal with.

91
00:12:06.480 --> 00:12:07.110
Josh Moore: Go for it. No.

92
00:12:07.500 --> 00:12:12.720
Laurent Guerard: Okay. Hi, everyone. I'm long and I'm working at the University of puzzle.

93
00:12:14.130 --> 00:12:23.250
Laurent Guerard: And where we are where we have our own own metal instance there and I'm more and I'm doing, you mentioned, this and that.

94
00:12:25.410 --> 00:12:35.520
Laurent Guerard: And learning that and I'm more your to know be kept in the loop about what is coming soon. And what is going on with the next generation file for

95
00:12:44.910 --> 00:12:53.010
Raphael Maree: No. Yes. Hello, everyone. I'm from the University of yes we are working on the site of mine and by a flow tools for

96
00:12:53.070 --> 00:12:54.690
Josh Moore: Sharing and analyzing and

97
00:12:54.900 --> 00:13:14.130
Raphael Maree: Looking machine learning and image analysis tools, as you know, we are working on using different file formats, including me. This became interested to know more about a MISA because of to know for iPads to call images we are using HDFS five dogs. That's

98
00:13:16.110 --> 00:13:16.530
That's it.

99
00:13:17.790 --> 00:13:18.060
Thanks.

100
00:13:20.100 --> 00:13:20.880
Robert Haase: Hello everyone.

101
00:13:21.210 --> 00:13:35.490
Robert Haase: My name is Robert. I'm from Germany. It looks like I become group leader quite soon in the shoes of big image data cloud something so I'm stronger the processing side and I'm here to learn what's happening on this story.

102
00:13:46.980 --> 00:13:47.310
Stephan Wagner-Conrad: Thank you.

103
00:13:48.540 --> 00:14:07.920
Stephan Wagner-Conrad: Stefan Wagner Conrad I'm working with Carl says microscopy. I'm a software architect working on his end platform and I'm here to learn about the next generation file formats and understand kind of the needs and and how to probably integrate this with our platform.

104
00:14:11.790 --> 00:14:24.390
Josh Moore: So thanks, everyone, for coming. So briefly to the oil me team. So I think there's at least one face. I don't know your so I'm Josh more from the only team. I'm actually based in Germany work for the University of Dundee

105
00:14:25.980 --> 00:14:29.280
Josh Moore: I ultimately all this is my fault. So you're welcome to play me for most of this

106
00:14:30.720 --> 00:14:34.920
Josh Moore: However, when it comes to funding, who we blame is Jason So I'll let him introduce himself.

107
00:14:36.300 --> 00:14:37.260
Josh Moore: Keeping all this going

108
00:14:37.440 --> 00:14:52.110
Jason Swedlow: Morning afternoon, evening or just swallowed University work with your team. Yeah, largely, my job is to keep everything funded and keep everything more or less on the rails. So we could do great work and it's

109
00:14:53.310 --> 00:14:55.560
Jason Swedlow: I think I'll just repeat what a lot of people said, I mean,

110
00:14:56.910 --> 00:15:03.510
Jason Swedlow: We're doing enough. A lot of work, but really interested in seeing what you guys are doing this direction because provides a lot of

111
00:15:06.210 --> 00:15:07.200
Jason Swedlow: Directions to go

112
00:15:13.710 --> 00:15:16.650
Josh Moore: Is the only team want to do their names alphabetically, David.

113
00:15:19.500 --> 00:15:25.920
David Gault: Yes, so I'm difficult, I'm primarily the bio formats developer for quite a number of years.

114
00:15:31.440 --> 00:15:32.310
Francis

115
00:15:33.660 --> 00:15:42.900
Frances Wong: Hi everyone I'm friends as well and I lead the idea project I'm I work for Sunday, but I'm actually based in Edinburgh.

116
00:15:47.310 --> 00:15:47.760
Frances Wong: Gym.

117
00:15:49.860 --> 00:15:52.800
jean-marie burel: So if you're working Sunday.

118
00:15:53.940 --> 00:16:10.350
jean-marie burel: So my focus at the moment is mainly analysis and building bridges between ultra to connect third party tools tomorrow. And we do that, maybe in the cloud. So for example, the goal for people interested in how to desert resort.

119
00:16:14.850 --> 00:16:15.180
Josh Moore: Peter

120
00:16:15.510 --> 00:16:26.370
Petr Walczysko: Hi, my name is better quality score I mainly with the outreach sessions and quality assurance and he keeps us honest.

121
00:16:27.420 --> 00:16:38.070
Sebastien Besson: said hi I'm Celeste and bustle also working from Edinburgh for me team has been been working around emerging formats from by formats to

122
00:16:38.730 --> 00:16:48.900
Sebastien Besson: Limited to everything that's coming to the idea or very interested in next generation file formats to scaling public resources of imaging and sensitive data.

123
00:16:50.910 --> 00:17:03.180
Simon Li: I am Simon Lee America development or so, one of the main idea, our developers. So I'm well aware of all the issues that I was running into with large data sets that hopefully this new format or so.

124
00:17:12.210 --> 00:17:20.730
Will Moore: Hi I'm will more and I'm mostly working on the client side the website of a mirror for the team.

125
00:17:22.260 --> 00:17:23.550
Will Moore: And yeah.

126
00:17:24.660 --> 00:17:25.590
Will Moore: Also bit of Python.

127
00:17:30.630 --> 00:17:32.460
Josh Moore: already joined. I don't know if

128
00:17:33.780 --> 00:17:34.620
I don't see him.

129
00:17:39.180 --> 00:17:40.890
Josh Moore: Yes, neither video or audio on

130
00:17:44.070 --> 00:17:47.010
Josh Moore: Oh, someone said, I forgot DOM too many D's

131
00:17:47.820 --> 00:17:49.080
Dominik Lindner: In time, so they get away with.

132
00:17:51.960 --> 00:18:07.320
Dominik Lindner: Dominic I'm working as part of the meeting Monday as well and they are usually I'm working. Getting data into it or and with respect to source and was mostly working recently on the CNI plugin to export.

133
00:18:08.580 --> 00:18:10.740
Dominik Lindner: Images from America to saw

134
00:18:14.850 --> 00:18:15.510
Your email.

135
00:18:17.610 --> 00:18:18.210
Oliver Biehlmaier: Hello.

136
00:18:19.350 --> 00:18:27.900
Oliver Biehlmaier: Sorry I'm late. I wasn't another call before. Um, yeah. So I'm only with you my I'm the head of the imaging core facility in Boston.

137
00:18:29.550 --> 00:18:31.470
Oliver Biehlmaier: And I just wanted to reconnect with you guys.

138
00:18:33.510 --> 00:18:37.380
Josh Moore: Could have you here. I think that's everyone I'm

139
00:18:39.330 --> 00:18:45.660
Josh Moore: To some extent. Now the chaos begins says to figure out how we just keep you know all these different balls in the air.

140
00:18:46.740 --> 00:18:49.590
Josh Moore: It sounds like so. So I tried to write down the

141
00:18:49.950 --> 00:18:54.480
Josh Moore: Three points that I heard everyone saying, and you know, I think the summary is

142
00:18:55.650 --> 00:19:03.180
Josh Moore: Are the thread that runs through everything is, what can we do to like reduce all of our overheads. Right. How can we do less work but get more out of it.

143
00:19:03.990 --> 00:19:21.690
Josh Moore: I'm not reproducing file formats, having easy to use plugins having analysis go faster. So that's certainly the goal. I assume it'll ultimately take years for us to get there. So pacing ourselves and just slowly making progress is what all this is about

144
00:19:22.920 --> 00:19:36.390
Josh Moore: Is there anyone who is kind of lost in general on what this is all about. So in terms of the format so bizarre got mentioned several times. I didn't make a video to what is our so I could understand if you know there's some

145
00:19:36.900 --> 00:19:48.870
Josh Moore: Just introductory kind of topics we need to cover or if anyone wants to talk about how we you're actually going to run this meeting. Like, what are we actually doing for the next hour and a half. And you just high level conversations, we need to have upfront.

146
00:19:51.720 --> 00:19:53.550
Christian Tischer: Maybe I have a high level question.

147
00:19:54.840 --> 00:19:59.280
Christian Tischer: So is om it's are the next generation file format.

148
00:20:00.720 --> 00:20:03.900
Christian Tischer: Is that an equivalent already or is that debate.

149
00:20:08.250 --> 00:20:14.100
Josh Moore: So I, I have to admit, I haven't been very clean with the definition. So that's probably something we could define

150
00:20:15.120 --> 00:20:27.600
Josh Moore: So initially, I think the next generation file for format and I'll probably start saying in GFS concept came from the progression TIFF HTML5 what's next.

151
00:20:28.230 --> 00:20:33.900
Josh Moore: And the reason I needed a new name is because there wasn't agreement yet on what we're doing on the format.

152
00:20:34.620 --> 00:20:41.940
Josh Moore: So this is some of what I said in my video. But in case, not everyone got to watch the, the couple of lightning talks that we put on put online.

153
00:20:42.810 --> 00:20:55.800
Josh Moore: There's been a process going on since the end of 2018 to unify at least two of the format. Um, I need my hands more so there's Tsar which is coming out of largely the Python space.

154
00:20:56.820 --> 00:21:07.680
Josh Moore: And then there's in five which is coming, largely out of the image J space. So Java space. They're both basically the same thing, which is a multi dimensional chunked

155
00:21:08.700 --> 00:21:17.640
Josh Moore: File format that can be stored well in the cloud, right. And so I think that's really the definition of in GFS that were most concerned with

156
00:21:19.680 --> 00:21:21.780
Josh Moore: There's an ongoing competition for to

157
00:21:22.980 --> 00:21:26.220
Josh Moore: Thank you. Stefan to, Who can find a good name so

158
00:21:27.720 --> 00:21:30.390
Josh Moore: For, you know, what is it we replaced that with

159
00:21:31.410 --> 00:21:36.060
Josh Moore: I don't feel like we have an agreement from the community that we can just call everything czar.

160
00:21:36.690 --> 00:21:45.180
Josh Moore: What we do have an agreement on is working together on a specification, at least a unifies are in five. So there's funding from season I

161
00:21:45.840 --> 00:21:59.850
Josh Moore: ongoing process. Hopefully by the end of the year, there will be a spec at least a preliminary version of the spec there's currently a request for comments so I can point people at that if you want to get involved in what that spec looks like now's the time. Um,

162
00:22:01.170 --> 00:22:14.760
Josh Moore: So I have some hope that by next year we'll have a single format that we can all start working with and then I'll deal with a naming issue. And it might be that in GFS gets replaced by czar and everyone's happy if everyone's not happy, then we'll figure out what to do about that.

163
00:22:16.260 --> 00:22:23.910
Josh Moore: The reason I'm not always quite clean about the naming is I think there's there other things that we're doing in terms of

164
00:22:25.560 --> 00:22:35.730
Josh Moore: Defining specifications that are valuable to other formats. You know, I think what we're doing in GFS. So in czar. Let me just use our you know what we're doing in the prototype in on these are

165
00:22:36.180 --> 00:22:42.390
Josh Moore: Can be applicable to HTML5 and maybe can even be applicable to TIFF, so I don't think those other formats are going away.

166
00:22:42.870 --> 00:22:46.530
Josh Moore: And there's going to be this nice suite of formats that all do what we want them to do.

167
00:22:47.220 --> 00:22:53.130
Josh Moore: Depending on how much funding, we get to keep everything moving forward. And how many people get involved and you know actually open P ours.

168
00:22:53.820 --> 00:23:05.250
Josh Moore: So I think it all works together so you can, you know, we can kind of cheat and say the whole bag will eventually be in GFS and everything will just be lovely and work and people won't have to care about you know reformatting their data, but

169
00:23:05.940 --> 00:23:16.530
Christian Tischer: So is, is it fair to say, just for my understanding that maybe all Emmys are is some kind of prototype implementation of the N G, F, F. Yes. Okay.

170
00:23:21.930 --> 00:23:22.500
Josh Moore: Cool.

171
00:23:23.700 --> 00:23:30.000
Josh Moore: Other high level questions. All right. Does anyone want to like rebuke like refute I just said. I mean, that doesn't have to be what's going on.

172
00:23:31.620 --> 00:23:37.260
Jason Swedlow: Especially if I one comment. Last, last night I had a two hour conversation with one of the UK funders.

173
00:23:37.680 --> 00:23:50.250
Jason Swedlow: It was sort of asking the same question that she just asked the way to, like, what's going to happen, right. I was very interesting. So, you know, these are people who are very well meaning a very serious have none of the technical knowledge.

174
00:23:51.600 --> 00:24:02.880
Jason Swedlow: But say, well, there has to be one format and you have to make one format and that will be everybody will use that. That's what we mean. Right. That would be the dream and sure that's a great dream.

175
00:24:03.360 --> 00:24:20.250
Jason Swedlow: And maybe that's actually a reasonable goal that's probably a separate question, but there's probably least what I was trying to communicate them is we are on a we're on a path and we're on a trajectory. And so I'm probably restating what you just said, which is

176
00:24:21.360 --> 00:24:30.540
Jason Swedlow: What we are currently calling on these are as a first step and then you'll say, well, okay, that's great. What's the second step, and that's probably why we're here to discuss that.

177
00:24:31.920 --> 00:24:34.320
Jason Swedlow: So I'm just repeating in a different way, but

178
00:24:35.310 --> 00:24:45.660
Sebastien Besson: Also, only one of the goal. I'll add here maybe compared to what happened before, is to involve very early, the community in these validation process.

179
00:24:46.710 --> 00:24:55.650
Sebastien Besson: We have done efforts of others have done it for a proposed solutions within their own group to build their own to solve their own problems like we all do.

180
00:24:56.310 --> 00:25:11.340
Sebastien Besson: But if want this thing to help them to be a reality. Actually, this kind of call will be necessary having a large number of players starting to Luton and define what they need and agree on something. So a real community effort.

181
00:25:17.250 --> 00:25:20.040
Josh Moore: Anyone else on the definition question.

182
00:25:26.340 --> 00:25:27.570
Josh Moore: Cool. Okay.

183
00:25:29.490 --> 00:25:29.850
Josh Moore: So,

184
00:25:30.150 --> 00:25:36.630
Josh Moore: I listed a couple of topics. Seems like several people are signing their names up to them. Currently, the order is somewhat random

185
00:25:38.100 --> 00:25:43.320
Josh Moore: I would hope that first topics have the most names on them. So at this point, if you kind of want to add your name to something

186
00:25:44.490 --> 00:25:54.600
Josh Moore: Or add a topic go about it. I assume everyone wants to stay together and we want to just kind of go through Topic and get as many points as we can.

187
00:25:55.350 --> 00:26:08.550
Josh Moore: I'm happy to set up breakouts, if that would be more interesting or maybe what we can do is, you know, start cutting people off. If we get too detailed and then say, towards the end, we can set up breakouts if necessary.

188
00:26:10.620 --> 00:26:18.540
Josh Moore: I figured, while people are adding things to the document what we could do is just have any questions for we don't have it well.

189
00:26:19.080 --> 00:26:27.510
Josh Moore: I can take questions for Trevor actually will can as well. So you have the videos that are up there is one posted this morning. Many of you haven't had much time to see it.

190
00:26:29.010 --> 00:26:41.670
Josh Moore: That's largely a JavaScript based one. So the underlying structure to infrastructure to what will showed there's wills Chris's presentation, which I'm sure a meal and Aaron can answer questions on

191
00:26:42.990 --> 00:26:45.630
Josh Moore: Are there any for the videos as they stand.

192
00:26:55.140 --> 00:26:55.950
Josh Moore: Alright, so I'm going to

193
00:26:57.000 --> 00:27:01.800
Josh Moore: Put everyone under pressure, who actually watch the videos are

194
00:27:04.170 --> 00:27:06.540
Josh Moore: Okay, so it looks like about 5050 right

195
00:27:07.920 --> 00:27:12.000
Josh Moore: All right. So it'll be interesting to get so towards the end, you know,

196
00:27:13.230 --> 00:27:19.800
Josh Moore: Either if we run over the two hours, whoever would like to stick around. So there's a there's a feedback part of all of this and figuring out

197
00:27:20.460 --> 00:27:27.300
Josh Moore: You know, are the videos useful, what would we actually need to do to make sure people watch them all of these kinds of questions are coming up so

198
00:27:27.990 --> 00:27:41.790
Josh Moore: Um, and then questions like, you know, does anyone want to film their own video to keep these conversations going but ok so I'm assuming we have no questions on the videos and we'll talk about the procedural whatnot Slater

199
00:27:43.170 --> 00:27:44.880
Josh Moore: Looking at the topics we have

200
00:27:48.120 --> 00:27:51.870
Josh Moore: Roy seem to be at the top of the list. Shall we just kick off Royce then

201
00:27:54.240 --> 00:27:57.390
Josh Moore: Okay, so I'll give you what I know.

202
00:27:58.800 --> 00:28:02.520
Josh Moore: A Mulan Aaron may have something to add to this. I know Eric has something to add to this.

203
00:28:03.990 --> 00:28:16.320
Josh Moore: If anyone wants to, like, okay, so can someone shut down the the voting list for Roy's and add who wants to talk and add questions and things in the notes. So from my side.

204
00:28:17.940 --> 00:28:35.760
Josh Moore: What we focused on to date are is the representation of masks in all Emmys, are we went through a number of different. I guess I can bring up the multi scale work. I probably won't be fast enough. I'm

205
00:28:37.320 --> 00:28:46.680
Josh Moore: Like a number of different representations. Those are all on GitHub issues that we can point to it if you want to, you know, kind of look at the history of what we tried out that process went throughout the summer.

206
00:28:47.370 --> 00:29:05.880
Josh Moore: What we ended up settling on was a five dimensional labeled image. So it's kind of your, your standard labeled image where each pixel represents a different label class, um, the big issue that we were trying to figure out is what to do with overlapping labels.

207
00:29:07.470 --> 00:29:16.890
Josh Moore: We kind of chose two solutions. Either you can have multiple labeled images in your own, these are you know and you can just you state that they're not overlapping

208
00:29:17.490 --> 00:29:33.060
Josh Moore: Or you can define a pixel value that is the overlap value, right. So that's as it currently stands, it's the proposals on image SC and it hasn't gotten much feedback I know one and dragon are working on it. So some feedback may be coming soon.

209
00:29:36.780 --> 00:29:43.290
Josh Moore: But I'd like to get much more. So we kind of put the label specification on hold and moved on to the high content screening specification, but

210
00:29:43.860 --> 00:29:48.750
Josh Moore: We are where we need to go back to it. We haven't done anything on the polygon front.

211
00:29:49.470 --> 00:30:04.050
Josh Moore: I know that's what so we need to define if we're talking about regions of interest or we only talking about polygons, or we also talking about masks from from the Dundee side, we've only been mask. I know a meal and Aaron have done some work on polygons. You guys want to

212
00:30:05.520 --> 00:30:07.020
Erin Diel: I mean no polygons in that

213
00:30:07.020 --> 00:30:21.180
Emil Rozbicki: Format right rather we focused on labels mosque Josh and implementation of that well implementation just just using whatever representation. This and putting it in work effectively and using in the web client and with the viewers.

214
00:30:21.780 --> 00:30:37.680
Emil Rozbicki: We looked at the polygons, but not in the format armies are, are, are any of the layouts, right, we were just looking at the possibilities of actually using the polygons as representations from us, but nothing else so far.

215
00:30:38.310 --> 00:30:42.930
Josh Moore: And so far. I mean, this was the general agreement, right, focusing on wk t and web

216
00:30:43.920 --> 00:30:57.630
Emil Rozbicki: At the moment, yes, that's what we use internally as well for our conversions for going from shapes to labeled masks as well. And yes wk BMW Katie seems to be what what we're using right now, for sure.

217
00:31:02.400 --> 00:31:23.190
Juan Nunez-Iglesias: Can I ask Josh kind of labeled implementation, does it let you apply the metadata to the classes. So if I want to say, you know, this class is nuclei. This classes mitochondria. Is there a place for that or is that still upcoming or is that out of scope.

218
00:31:24.450 --> 00:31:33.750
Emil Rozbicki: So, who on what we've done. If you look at Chris's video we've actually show how that works on our part, but it's split. It's not in the file itself at the moment.

219
00:31:34.620 --> 00:31:43.800
Emil Rozbicki: I don't think there's any specification for that for that we've actually started the features that are linked to the particular pixel volume in America table.

220
00:31:44.730 --> 00:31:57.300
Emil Rozbicki: And if you look at the video, the viewer is actually doing the resolution. So the viewers are asking server. Hey, what are the values for despotic particular pixels. So I can apply the particular rendering are filtering in the client so

221
00:31:58.110 --> 00:32:07.140
Juan Nunez-Iglesias: Yeah. And if we were to say is, is there a scope for putting it into into those editors at some point or not.

222
00:32:07.410 --> 00:32:10.860
Emil Rozbicki: Would be definitely interesting for us as well. But we haven't actually

223
00:32:10.920 --> 00:32:12.210
Emil Rozbicki: decided on the format yet.

224
00:32:13.020 --> 00:32:15.540
Juan Nunez-Iglesias: So, yeah, yeah. So we might

225
00:32:15.960 --> 00:32:18.120
Juan Nunez-Iglesias: That we sort of want that. So we might come in.

226
00:32:21.420 --> 00:32:22.080
Josh Moore: Yeah, and

227
00:32:22.650 --> 00:32:24.120
Josh Moore: I guess, to some degree, that would

228
00:32:24.390 --> 00:32:35.760
Josh Moore: Is exactly what we would like to get out of this. So from our side, having people come in and say I want X. I'm going to open a PR against the spec and the library and it just kind of appears, would you wonderful

229
00:32:36.540 --> 00:32:43.410
Josh Moore: If anyone needs a walk through and how to do that, you know, we can certainly put that together, either as a video or whatever.

230
00:32:45.060 --> 00:32:49.260
Eric Perlman: So they talk briefly about the polygons. So I like the idea of the

231
00:32:49.860 --> 00:32:56.220
Eric Perlman: Well known binary or well known text for polygon definitions, but it seems like we're sort of in to split worlds here. There's one where

232
00:32:56.430 --> 00:33:05.520
Eric Perlman: It's a file, you're going to export describing annotated regions and masks and then there's another where I think we're me and Dave currently are which is we have data stored in Romero

233
00:33:05.970 --> 00:33:12.390
Eric Perlman: And we're, you know, interacting with ROI is there to find an ROI. So it's very sort of read write basis, not an export quality.

234
00:33:13.020 --> 00:33:27.360
Eric Perlman: Yet we're using those are wise to access the czar's that's where my interest in terms of, like, okay, we have this polygon definition, but now I want to read it at one eight scale. Let's open up the Tsar file read that resolution translate the coordinate to get the images out

235
00:33:27.600 --> 00:33:34.830
Eric Perlman: So it's sort of in that description there, we are still storing the tables in America. It seems like you guys are also sort of in this world where sometimes he's span.

236
00:33:35.400 --> 00:33:43.050
Eric Perlman: America access made the data and this new file format for images, but then you only want to share with the world you combine them together and I just say

237
00:33:43.110 --> 00:33:51.630
Eric Perlman: I don't really have a concise statement here, but it's just something might have to kind of going back and forth about where data should live based on what sort of stage it's in

238
00:33:53.730 --> 00:33:54.510
Josh Moore: Yeah, so I mean

239
00:33:55.620 --> 00:33:58.530
Josh Moore: Romero's view of the world is

240
00:34:00.240 --> 00:34:00.990
Josh Moore: You know, it could

241
00:34:02.880 --> 00:34:20.610
Josh Moore: Um, and especially so with some types of data, the tabular data, the region of interest data. And I think we've we've run into limits that we would either need to start re architected Romero internally to figure out a better way to store things build the API, etc, etc.

242
00:34:22.170 --> 00:34:27.510
Josh Moore: Or basically loosen that ownership and say, and there's going to be a barrier somewhere in the data graph.

243
00:34:27.930 --> 00:34:35.250
Josh Moore: You have your images and it's all in the marrow and then somewhere there's a link to something else. And, you know, we've worked through some

244
00:34:35.730 --> 00:34:47.490
Josh Moore: Is like microservices. The Tsar work is another place that we, you know, another store that we can put the data. Um, we're not to the point with the army's our prototype that we've

245
00:34:49.500 --> 00:34:54.090
Josh Moore: Yeah, engaged with Romero to take all the ROI is out. Right, so they would have to be a process to

246
00:34:54.480 --> 00:35:02.190
Josh Moore: extract them delete them and store them elsewhere, but one more thing, I think, I think, though, that's where we want to go. Is that the client.

247
00:35:02.610 --> 00:35:12.240
Josh Moore: Is talking to a Miro but also loading other types of data somewhere else and that would bring the unification of the two worlds, you're talking about, you know, so if

248
00:35:13.050 --> 00:35:22.500
Josh Moore: The Roy's are stored in this format that we're all talking about here, and that's where you're loading them from then it's only one place, it is already they exported format. Right.

249
00:35:23.760 --> 00:35:27.360
Eric Perlman: Oh yeah, I mean, that's the sort of workflow, we're doing right now and works quite well.

250
00:35:28.380 --> 00:35:43.140
Eric Perlman: I guess I would also like to say if we have the ROI is and masks be kind of nice to logically be able to have them separated from the pixel data as an option, maybe not as a requirement because I can imagine many workflows, where we have, you know, several terabytes of slide data, but

251
00:35:44.190 --> 00:35:53.430
Eric Perlman: Someone changes the ROI is for their use case, it might be nice to be able to say ROI is from this this file, combined with pixels from that file, rather than having to be the same one.

252
00:35:53.910 --> 00:36:01.080
Eric Perlman: I don't see that as a major complication, but just that that'd be one way of handling rewrite, if you will, on large image data.

253
00:36:04.470 --> 00:36:07.350
Juan Nunez-Iglesias: That already works with bizarre. Is that, is that right

254
00:36:08.730 --> 00:36:15.960
Josh Moore: So one certainly so I'll just throw you under the bus one he's currently shipping the satellite imagery in czar.

255
00:36:16.020 --> 00:36:16.860
Josh Moore: Just kind of cool.

256
00:36:18.990 --> 00:36:22.530
Josh Moore: It's on our servers. But yeah, we don't tell people about it. Anyway, um,

257
00:36:22.950 --> 00:36:34.410
Josh Moore: And yeah, so you can have a file that is just the labels for example that's completely fine. Um, I think, I think what we're missing and we touched on it. Oh.

258
00:36:35.460 --> 00:36:44.490
Josh Moore: Stefan dropped, um, you know, we touched on it briefly is, is basically a way in the specification to have a remote link right and the data.

259
00:36:45.120 --> 00:36:54.150
Josh Moore: And it's, you know, completely obvious that you need to have that. It's just, it'd be good to get it right so that it works for a large number of use cases.

260
00:36:54.570 --> 00:36:59.970
Josh Moore: So that may actually be something we want to talk about is what are all the ways that you want to have some data and Azhar

261
00:37:00.540 --> 00:37:15.930
Josh Moore: And you want to be able to navigate to the originals are or a child's are right. So we kind of punted on that and said, We'll deal with that as a community when the time comes. If you guys are saying, the time has come, and we need it now. Okay, fair enough. But

262
00:37:19.050 --> 00:37:20.130
Juan Nunez-Iglesias: I've been using

263
00:37:20.310 --> 00:37:24.900
Juan Nunez-Iglesias: LinkedIn my head for the past decade, so I can keep using it for another year.

264
00:37:27.270 --> 00:37:36.000
Eric Perlman: I definitely like that idea. I mean, particularly if links can be relatives who have the ability of like choosing which mere or word source you're pulling them from

265
00:37:36.480 --> 00:37:45.870
Eric Perlman: Because I definitely know world where. Yeah, if I have the data locally on desk, because I'm rapidly working with something that's great. And then I'm going to publish it to stationary S3 bucket. It'd be nice to be able to just copy it.

266
00:37:46.560 --> 00:37:50.070
Eric Perlman: But that I think that would solve all the issues I was just describing

267
00:37:57.810 --> 00:38:05.520
Juan Nunez-Iglesias: independent labels work now is that does that get you started. Eric, or are you gonna wait for the links.

268
00:38:06.420 --> 00:38:12.000
Eric Perlman: What we're doing, I mean our label file would be terabytes and size on, you know, on a 10 terabyte

269
00:38:12.030 --> 00:38:14.760
Juan Nunez-Iglesias: Data set, but they labels labels compressed really nicely.

270
00:38:15.870 --> 00:38:19.320
Eric Perlman: They are so much lower than using rectangle ROI. So

271
00:38:20.760 --> 00:38:23.910
Eric Perlman: You are wise is to be able to know what needed to be

272
00:38:24.390 --> 00:38:34.500
Eric Perlman: And so out of like 500 sections on slides. I want to some coordinates of pixels three labeled data. Yeah, you have some optimizations there you have your indices and things like that. It's

273
00:38:35.130 --> 00:38:47.970
Eric Perlman: Gonna be my God, yeah, I'm fine. Right now using reading the polygons Romero and I like the idea that that can transition to using wk t from, you know, attribute files or something like that.

274
00:38:52.290 --> 00:38:55.980
Josh Moore: Anyone who hasn't spoken up on polygons. Wanna get their two cents and

275
00:38:57.570 --> 00:38:57.840
Will Moore: What is

276
00:38:58.320 --> 00:38:58.620
It.

277
00:39:00.030 --> 00:39:01.320
Josh Moore: Well known text.

278
00:39:01.410 --> 00:39:01.980
Will Moore: So it's a

279
00:39:02.220 --> 00:39:03.060
Josh Moore: Polygon format.

280
00:39:07.890 --> 00:39:12.990
Eric Perlman: From the geospatial world Devon was every definition to select the region of space.

281
00:39:16.380 --> 00:39:18.660
Eric Perlman: And also supported by all the major database.

282
00:39:21.240 --> 00:39:23.700
Anatole Chessel: So anyway, we could want.

283
00:39:24.180 --> 00:39:34.050
Anatole Chessel: Several comments, so I know for the body in the yoga world, but you see is the traces. So look up the trees.

284
00:39:35.310 --> 00:39:54.840
Anatole Chessel: I can imagine other even from Michigan pretty good other formats that would some people could use so how how we how would we be possible to have a several, several like internal ways, things are included America they are tagged as being one on the other. And that would be

285
00:39:55.170 --> 00:39:55.470
Josh Moore: Created

286
00:39:55.650 --> 00:39:57.690
Anatole Chessel: On the fly drones.

287
00:39:58.680 --> 00:40:06.570
Josh Moore: Yeah, I mean ultimately anything's possible right i mean we can have as many micro formats as we want. But the problem is supporting it long term.

288
00:40:09.780 --> 00:40:21.390
Emil Rozbicki: And probably building queries on the top of all of those specifications. Right, so you will need a unifying format somewhere, anywhere because you want to optimize how the server will retrieve those things and show them the images.

289
00:40:22.620 --> 00:40:35.460
Anatole Chessel: And that's another thing. Well, that just, it could be just been a little bit of a encoded like having the right, the right pipeline to encode in the whatever format is is a truism.

290
00:40:36.690 --> 00:40:40.590
Anatole Chessel: We just need that format to be flexible enough to handle any cases.

291
00:40:41.610 --> 00:40:52.650
Anatole Chessel: So one. One thing I would like when wondering, I don't know about who he is, how much can you encode as well that are that that leaves on the mesh. So, but whatever.

292
00:40:53.880 --> 00:40:54.510
Anatole Chessel: You do the

293
00:40:55.530 --> 00:40:56.940
Anatole Chessel: Measurements you have

294
00:40:56.970 --> 00:40:59.700
Emil Rozbicki: W wk t just specify vertices of the

295
00:40:59.700 --> 00:41:06.270
Emil Rozbicki: Polygon there is no meta data associated with anything in that format, tomato, it's plain text for practices.

296
00:41:07.320 --> 00:41:08.040
Emil Rozbicki: That's all it does.

297
00:41:09.600 --> 00:41:12.090
Emil Rozbicki: So everything else need to be linked on the top of that somehow

298
00:41:14.010 --> 00:41:15.810
Anatole Chessel: So how you definitely got that.

299
00:41:16.890 --> 00:41:17.310
Anatole Chessel: We would want

300
00:41:18.750 --> 00:41:20.520
Anatole Chessel: Because that would be convenient but

301
00:41:21.990 --> 00:41:31.380
Emil Rozbicki: That's that's the direction where we want to go. And that's what we're thinking about currency right I'll to actually link those four months with the features and other things. Yes.

302
00:41:32.490 --> 00:41:33.570
Anatole Chessel: But then would

303
00:41:34.980 --> 00:41:44.910
Anatole Chessel: Maybe there's other format than the Wi Fi that have that real team where we could encode biometric stuff on top of the policies are right there.

304
00:41:45.150 --> 00:41:49.980
Emil Rozbicki: And i mean i i totally agree, we are looking at w Katie and wk be because

305
00:41:50.010 --> 00:41:54.630
Emil Rozbicki: As Josh said this is not only well known. It's also well supported

306
00:41:55.710 --> 00:42:12.660
Emil Rozbicki: With other farmers, the knots, not really given especially all the major database engines to support W k t and wk be kind of natively and you can build spatial indexes based on those which was pretty nicely extends all the basic functionalities, we have in the viewers as well.

307
00:42:13.920 --> 00:42:14.400
Sebastien Besson: And I told

308
00:42:16.200 --> 00:42:16.860
Sebastien Besson: Whatever you

309
00:42:18.150 --> 00:42:19.500
Josh Moore: Are good just keep talking now. Yeah.

310
00:42:21.270 --> 00:42:30.540
Josh Moore: If you have experienced with a format that you know will do this, then you know that's something we're here to do as well. Let's gather that knowledge. So that's roughly what you were saying as well so

311
00:42:30.690 --> 00:42:42.510
Sebastien Besson: Yeah, I was. I was going to say. I mean, do you mean use it for these various some another one that you have in mind, specifically that would come from well and community that would be, you know, what would be the candidates for to be

312
00:42:42.510 --> 00:42:43.470
Sebastien Besson: Thinking of right now.

313
00:42:43.500 --> 00:42:44.940
Anatole Chessel: So the one thing off.

314
00:42:46.230 --> 00:42:54.480
Anatole Chessel: Other if there's anybody else in the coal that deals with no traces but WC which means my web

315
00:42:55.830 --> 00:42:59.400
Anatole Chessel: Which is another plain text simple thing.

316
00:43:00.600 --> 00:43:06.480
Anatole Chessel: Which codes, no traces from connector MC cities.

317
00:43:08.520 --> 00:43:11.580
Josh Moore: There's already a PR for a party for it. Sorry.

318
00:43:13.500 --> 00:43:15.930
Anatole Chessel: Yeah i would i would assume would be a sort of thing because

319
00:43:17.700 --> 00:43:18.720
To deal with

320
00:43:20.010 --> 00:43:31.590
Anatole Chessel: So nothing Graham said that you want with working with the support from neuroscience and i would i would assume that this would be from one of the other thing that would meet

321
00:43:39.600 --> 00:43:43.110
Josh Moore: Did you wanna just generally thumbs up or you want to add anything

322
00:43:44.520 --> 00:43:45.570
Josh Moore: Okay, fair enough.

323
00:43:47.100 --> 00:43:48.510
Josh Moore: Tissue you were gonna say something a bit ago

324
00:43:48.870 --> 00:44:05.100
Christian Tischer: Yeah, no, I just wanted to clarify or maybe what could you answer it. So like what we mean with meta data is it like what I would call features, maybe, so you have to label them as image and then for each label you have stuff like a name or class or some measure

325
00:44:05.160 --> 00:44:16.590
Christian Tischer: Value. Okay, so, and is this a tabular format. Good enough. Is that or are you guys thinking you need something more complex maybe

326
00:44:16.920 --> 00:44:34.110
Juan Nunez-Iglesias: High in our use cases just is our satellite imagery and this is whether its water or urban landscape or forest or whatever. So it's just that the labels, the segmented labels has a specific meaning as opposed to having like you know different

327
00:44:34.200 --> 00:44:34.680
Different

328
00:44:35.760 --> 00:44:46.560
Christian Tischer: Okay, but then the table would do when you are good, just the one column table. Correct. Right. Okay. Because I will need more columns, but then maybe I could agree.

329
00:44:49.260 --> 00:44:49.470
Juan Nunez-Iglesias: With

330
00:44:51.150 --> 00:44:51.570
Juan Nunez-Iglesias: That too.

331
00:44:54.390 --> 00:45:03.000
Christian Tischer: But OK and said something just for you. I think I know what are you thinking about getting a table into is our right.

332
00:45:03.810 --> 00:45:12.930
Josh Moore: Yeah, so I ran into a blocker. I was going to add it to the initial proposal, um, the, the blocker actually came on the Java side of things. So

333
00:45:13.860 --> 00:45:23.940
Josh Moore: They're very real barriers at the moment just to make keep things working in both Java and Python and you know. So to some extent we're kind of the least common denominator.

334
00:45:24.270 --> 00:45:29.340
Josh Moore: And if we can get if we can come to all these agreements and use a single format and update all the libraries.

335
00:45:29.760 --> 00:45:37.530
Josh Moore: Then I think we can start really pushing on some of these things. The problem was in in five if you encode, you know, a stringed a

336
00:45:38.130 --> 00:45:43.320
Josh Moore: Basically a CSV into the Tsar job. I can't read it anymore. I'm

337
00:45:44.250 --> 00:46:01.650
Josh Moore: On the Python side it was fairly straightforward. So I think having metadata, at least on the labeled image would be a very easy win. And if someone can help us with in five, we can fix that. I'm for the vertices, you know, if someone has a format where the table can be linked

338
00:46:03.420 --> 00:46:10.620
Josh Moore: Additionally, you know, so you have your core format and finding the mesh the vertices have some kind of object ID and that goes to a table.

339
00:46:11.190 --> 00:46:22.560
Josh Moore: Then we could link everything together. Right. So the segmentation good link to the table, the mesh good link to the table, you know, maybe that's not necessary, but I think with a couple of fairly simple primitives.

340
00:46:23.940 --> 00:46:33.510
Josh Moore: Labeled images polygons meshes and tabular data, we actually get a lot of the use cases that all of us need basically right away.

341
00:46:36.990 --> 00:46:37.260
And

342
00:46:38.340 --> 00:46:45.510
Guillaume Gay: How about simpler primitives for our eyes and like circles, rectangles. Is it

343
00:46:47.550 --> 00:46:53.250
Josh Moore: So the specs are so the one the XML on the TIFF spec certainly has those, I guess.

344
00:46:54.780 --> 00:47:00.690
Josh Moore: Leave the question to everyone else, because the simpler, we can keep things the fewer things we needed to find the better, but

345
00:47:03.420 --> 00:47:04.050
Josh Moore: Circles.

346
00:47:05.070 --> 00:47:11.250
Josh Moore: Spheres David was no one circles polygons everywhere. And I tell us know and

347
00:47:12.960 --> 00:47:24.510
Jean-Karim Heriche: So the all these. They're basically geometric shapes that are based on the defined, but mostly defined by an anchor point so

348
00:47:25.080 --> 00:47:38.370
Jean-Karim Heriche: You could just have something flexible where there is an anchor point and an associated distance threshold distance. So you could decide the Caesar, the side of a cue ball this other radius of a sphere.

349
00:47:40.740 --> 00:47:45.390
Jean-Karim Heriche: And you can encode a number of shapes in this way, if that's what

350
00:47:47.340 --> 00:47:59.520
Jean-Karim Heriche: Is needed. I think the main the main use case here is really the the dependence on on an anchor point to do some our center of that our I

351
00:48:04.710 --> 00:48:06.630
Josh Moore: I mean certainly having points may make sense.

352
00:48:06.630 --> 00:48:07.620
Jean-Karim Heriche: For just

353
00:48:07.830 --> 00:48:08.910
Josh Moore: Defining something right

354
00:48:13.920 --> 00:48:16.680
Josh Moore: Okay, so it says, thinking back summarizing

355
00:48:18.810 --> 00:48:24.660
Josh Moore: No immediate objections to what went on with the label specs. I assume that's why I got no feedback. So I'll take that as a positive

356
00:48:26.130 --> 00:48:41.820
Josh Moore: Need to think about the one or more specifications for Roy's that we're going to talk about so sorry for polygons minimally and possibly other things and critically linking that to metadata, but no one at the moment has a clear solution IE. It's investigation right

357
00:48:46.290 --> 00:48:46.680
Josh Moore: Cool.

358
00:48:47.010 --> 00:48:53.340
jean-marie burel: I think for our oil is the two world of the political world where you will have

359
00:48:54.720 --> 00:49:08.520
jean-marie burel: You will probably will need circle. But if you start drawing things if you can user interaction, then you start to have standard cheap that you may not want to present with poverty line polygon or anything like that.

360
00:49:09.600 --> 00:49:10.560
So I think that is a

361
00:49:11.730 --> 00:49:12.270
Josh Moore: Okay, so

362
00:49:13.980 --> 00:49:18.900
Josh Moore: Possible argument for having a more complex language. Okay, fair enough.

363
00:49:19.110 --> 00:49:21.090
jean-marie burel: I think it depends from which angle you come from.

364
00:49:22.890 --> 00:49:23.490
Josh Moore: David, go for it.

365
00:49:24.180 --> 00:49:27.630
David Pinto: Yeah, so I just, I mean, the reason I say no one circles easy like

366
00:49:28.080 --> 00:49:37.050
David Pinto: Every time I i've seen done is just because just easy. People are select circle. Draw a circle, but really what we want to do is, you know, select an area that is roughly circular

367
00:49:37.860 --> 00:49:41.940
David Pinto: The reason why they put circles, not because they need a circle is because it's just easier to draw

368
00:49:42.330 --> 00:49:57.450
David Pinto: So, I mean, if we have a format that doesn't support circles. By default, and you're just adding it so make it more complicated for us to support a case after doesn't exist. It's just, you know, like ease or drawing. I don't think it's not. I don't see a need for it.

369
00:49:57.780 --> 00:50:02.340
David Pinto: It's an invite in biology, things are never nicely.

370
00:50:02.400 --> 00:50:16.830
David Pinto: You know geometrical shape you know they're not liking the things are not square like roughly around that or like nothing never perfectly circular, I think, you know, made a point about like the point in that should be enough. You don't need a radius around that point.

371
00:50:18.390 --> 00:50:24.810
David Pinto: So a surface. You know, people just take that that that that and it's roughly around circling cells are never perfectly circle.

372
00:50:26.190 --> 00:50:26.430
Oh,

373
00:50:29.070 --> 00:50:29.460
David Pinto: Cool.

374
00:50:30.120 --> 00:50:31.680
Josh Moore: So yeah, it sounds like

375
00:50:31.740 --> 00:50:44.040
Josh Moore: This is a conversation that's going to keep going. You know, I assume, lots of that's an image. See, we can certainly write up what's here. Get it posted have other opinions. Who knows what will happen this afternoon. Right.

376
00:50:45.780 --> 00:50:46.140
Anatole Chessel: So me the

377
00:50:46.260 --> 00:50:47.550
Anatole Chessel: One thing I can, I can add

378
00:50:48.120 --> 00:50:58.170
Anatole Chessel: On that is that one thing that would not be included in the polygons is if you want actually continues objects. So what it says plans or things like that.

379
00:50:59.760 --> 00:51:07.260
Anatole Chessel: So, and that we don't go chuckles and any other show that you want, but then you would have a distinction between discrete

380
00:51:08.310 --> 00:51:22.920
Anatole Chessel: Good sized objects which are facets and not really a couple seasons and the services and well as clients and couldn't use nerves, or whatever, which might, you might want to encode and might not be in the

381
00:51:24.810 --> 00:51:25.140
Anatole Chessel: World.

382
00:51:26.790 --> 00:51:28.110
Josh Moore: It's a good point. So certainly

383
00:51:28.740 --> 00:51:29.460
Anatole Chessel: Vision, the

384
00:51:29.610 --> 00:51:36.810
Josh Moore: Allman nimble API is working on that and talk to us about, you know, how do I go about sharing what i'm doing so.

385
00:51:37.620 --> 00:51:44.160
Josh Moore: We should definitely get her on board. And I don't know if any of the any of you are currently storing splines for what you're, you know,

386
00:51:44.550 --> 00:51:51.120
Josh Moore: Recording and your images. If you have examples. As always, we need to see the data. What does it look like and then we can talk about what the format would be

387
00:51:52.920 --> 00:51:56.820
Josh Moore: Okay. Any last comments otherwise slowly shifting gears here.

388
00:52:01.710 --> 00:52:03.240
Josh Moore: Anyone need to stretch. Yeah.

389
00:52:04.410 --> 00:52:15.660
Josh Moore: I can do a round of yoga had that had that recently at a meeting. Um, it was good. That's good. Um, cloud is next, put it in quotes because

390
00:52:16.500 --> 00:52:16.770
Jason Swedlow: Right.

391
00:52:16.800 --> 00:52:17.910
Josh Moore: The cloud is just someone else's

392
00:52:17.910 --> 00:52:18.570
Computers.

393
00:52:22.680 --> 00:52:25.950
Josh Moore: Does everyone have an innate feeling for what the cloud is what it means.

394
00:52:29.580 --> 00:52:30.810
Jean-Karim Heriche: Well, I could

395
00:52:31.950 --> 00:52:39.240
Jean-Karim Heriche: Just say quickly what we what we see coming in the context of of your life.

396
00:52:39.990 --> 00:52:40.290
Sure.

397
00:52:42.420 --> 00:52:43.260
Jean-Karim Heriche: So,

398
00:52:44.430 --> 00:52:48.720
Jean-Karim Heriche: Basically, the, the, the idea is that you

399
00:52:49.680 --> 00:53:06.540
Jean-Karim Heriche: Your data will leave somewhere and be publicly accessible and and that basically funders will not keep on buying you more computers and data centers, but will give you credits to compute somewhere else in big data centers so

400
00:53:07.860 --> 00:53:23.010
Jean-Karim Heriche: How do we move from the current model of working out of workstations local workstations, or local H PCs, for those that ask them to working on images remotely.

401
00:53:24.030 --> 00:53:25.110
Jean-Karim Heriche: That's the

402
00:53:26.160 --> 00:53:29.130
Jean-Karim Heriche: Kind of main challenge that I see.

403
00:53:32.550 --> 00:53:37.770
Jean-Karim Heriche: The example I take for is the neural imaging center, we are building in either bag.

404
00:53:39.120 --> 00:53:51.180
Jean-Karim Heriche: At the moment, the current model for visitors at a table is that they get basically a Unix account in a demo. So they basically are virtual employees.

405
00:53:52.020 --> 00:54:04.320
Jean-Karim Heriche: But with a facility now going to reach in the hundreds of users of visitors. We're going to be very soon, adding more virtual employees. Then we all employees.

406
00:54:04.980 --> 00:54:17.220
Jean-Karim Heriche: And and the end those people will not stay at the amble workstations for for the six months or years that their projects or data analysis will need so they will have to

407
00:54:17.220 --> 00:54:17.970
Jean-Karim Heriche: Go back.

408
00:54:18.300 --> 00:54:27.510
Jean-Karim Heriche: To their home institution and still be able to access their search, eat your bytes of data because very often they won't affect the mean of dealing with it back room.

409
00:54:29.550 --> 00:54:47.340
Jean-Karim Heriche: And so this is kind of one examples where basically you visit the facility you produce a massive amount of data that you can't process back home, but the facility you visited actually as a cloud where you can actually do that work.

410
00:54:53.550 --> 00:54:55.380
Jean-Karim Heriche: So this anymore.

411
00:54:56.490 --> 00:55:00.810
Jean-Karim Heriche: Which was honing those the computer basically

412
00:55:01.920 --> 00:55:11.430
Simon Li: Okay. Sean cream from the perspective of the data format that we're discussing today. What, what kind of key requirements. Do you think they are now. We haven't really covered

413
00:55:12.450 --> 00:55:13.560
Simon Li: S3 storage, for example.

414
00:55:13.620 --> 00:55:14.280
Jean-Karim Heriche: Yeah, so that

415
00:55:14.550 --> 00:55:15.270
Simon Li: We think about

416
00:55:15.450 --> 00:55:15.810
Jean-Karim Heriche: Yeah.

417
00:55:15.900 --> 00:55:17.880
Jean-Karim Heriche: So I think it was written somewhere.

418
00:55:18.510 --> 00:55:24.300
Jean-Karim Heriche: Further down in the in the back end. It's about the granularity of data access

419
00:55:25.530 --> 00:55:26.280
Jean-Karim Heriche: So,

420
00:55:27.510 --> 00:55:36.180
Jean-Karim Heriche: Can I over HTTP extract a particular sub volume, for example, for processing.

421
00:55:38.010 --> 00:55:54.480
Jean-Karim Heriche: In my brother. So if I mean at the moment you saw something in in in one bucket and for most of the of the current tools you have to download this old thing and actually copy to a file so that you can process it.

422
00:55:56.880 --> 00:56:08.910
Jean-Karim Heriche: So, now can I can I access smaller amount of data for either processing and distributed processing or for visualization

423
00:56:12.480 --> 00:56:25.200
jean-marie burel: I think that's the type of example we've been looking into shock Kareem is actually you you you retrieve the you know the structure of dissolve the plate, for example.

424
00:56:26.160 --> 00:56:37.020
jean-marie burel: And you know exactly which channel. You want to analyze and which which field of view you want and which channel you can actually request that and it makes things much much more easier, much easier.

425
00:56:37.470 --> 00:56:46.050
jean-marie burel: And also the advantages. You can set up and we have example of that. You can start doing distributed element using task.

426
00:56:48.990 --> 00:56:58.440
jean-marie burel: The task library and you can do a schedule. What is going to up and you prepare the function you want and then you you process all in parallel, which is much nicer than

427
00:56:59.490 --> 00:57:00.630
jean-marie burel: Than what we have done.

428
00:57:02.490 --> 00:57:18.060
jean-marie burel: What is currently possible even with the, the current API of idea where you retrieve plane and because of the nature you it's quite it's a bit more complex to distribute things and in that case is kind of I would say pretty straightforward with our

429
00:57:20.130 --> 00:57:37.320
jean-marie burel: And you only fetch what you need. Basically, and this quarter and you, you, you, you, you retrieve a shell, which is kind of it was a call it the dissolve set up and then you have things organized and when you request it to binary that's much, much nicer. So, since that is

430
00:57:38.880 --> 00:57:45.330
jean-marie burel: Enabling I would say much more parallel computing and remote remote fetching what you need.

431
00:57:47.520 --> 00:57:52.320
Jean-Karim Heriche: Now, this is basically the kind of thing I want to know and be able

432
00:57:53.400 --> 00:57:57.630
Jean-Karim Heriche: To use all that in terms of the specification. That's kind of more what I wanted.

433
00:57:57.630 --> 00:57:57.990
Simon Li: To

434
00:57:58.650 --> 00:58:02.160
Jean-Karim Heriche: make people aware of that. Maybe this is something that can be

435
00:58:03.960 --> 00:58:04.590
Jean-Karim Heriche: Considered

436
00:58:05.610 --> 00:58:21.450
Simon Li: Okay so specifications be now obviously since our only meeting back in whenever it was it May or June, we've iterated on it since. Do you have any sort of direct feedback as to what the blockers are for you to say use it in terms of distributed computing remote access and so on.

437
00:58:22.290 --> 00:58:22.740
Simon Li: Well, at

438
00:58:22.890 --> 00:58:24.330
Jean-Karim Heriche: The moment I mean

439
00:58:24.510 --> 00:58:25.140
Jean-Karim Heriche: Probably I'm

440
00:58:26.400 --> 00:58:47.100
Jean-Karim Heriche: Behind on on on latest developments, but basically it's kind of the, the lack of tools and really what I would want to be able to do is really have images stored in in S3 buckets, though, that facilitates the sharing also because it easier for controlling access

441
00:58:48.210 --> 00:58:50.160
Jean-Karim Heriche: Because we can't imagine that you don't want to make

442
00:58:50.160 --> 00:58:54.660
Jean-Karim Heriche: Everything public and so you want to suit to be able to control access

443
00:58:55.830 --> 00:59:03.990
Josh Moore: So to give everyone a bit of a feeling for the weird thing stand. Currently, so. So in the end, you know, there are

444
00:59:04.800 --> 00:59:17.520
Josh Moore: I think it's five or six data sets that we've linked to. But, you know, oh, do we have terabyte. Yes, we definitely have terabytes of data that have been converted into the into the OMB. So our format and put up on S3 so EBS S3.

445
00:59:18.150 --> 00:59:22.050
Josh Moore: And you're welcome to download it and play with it, several of you have already done that.

446
00:59:23.580 --> 00:59:29.220
Josh Moore: In terms of this mean what john cream says is critical. I mean, so go back

447
00:59:30.270 --> 00:59:45.450
Josh Moore: Over a decade. Oh marrow allowed you to do this right. I mean, you could put an arrow somewhere, you could, you know, upload your images do it and you could access it from API. The downside is you must write your clients to access this you know custom API. It's

448
00:59:46.500 --> 00:59:56.430
Josh Moore: It's the problem. All of us have when you're building something new. So certainly part of what we're trying to achieve here is is an agreement from the community that we want to do something.

449
00:59:57.810 --> 01:00:11.760
Josh Moore: That we can all work together on so that you know we we clearly have to do the work at least once. But then, as I think as one was saying, you know, you don't have to worry about it ever again. It's just you can access it and I'll pull up a slide. Real quick, um,

450
01:00:12.840 --> 01:00:19.350
Josh Moore: And my feeling is at the moment that actually works pretty well. Um,

451
01:00:20.520 --> 01:00:28.170
Josh Moore: Part of the reason why we chose our is because there were lots of other people working on, you know, the Python libraries and the

452
01:00:29.550 --> 01:00:41.580
Josh Moore: The parallelization with das because Sean Murray mentioned and the another library library called FS spec, which allows you to transparently access either Google Cloud or, you know,

453
01:00:42.390 --> 01:00:52.860
Josh Moore: AWS S3 or basically, it all comes down to is you. You have a URL you give it to the library and you can access the data in parallel. So it's really exactly what you're talking about. JOHN cream.

454
01:00:53.490 --> 01:01:03.600
Josh Moore: Um, I think you can do so kind of answering Simon's question. I think you can do what you want. Today, the big caveat is um

455
01:01:06.540 --> 01:01:19.590
Josh Moore: Is it's not completely agreed on, right. So that's where we're trying to get with these community calls is saying, you know, are we all working in the same direction, or we all willing to start trends forming our data I'm into this format.

456
01:01:20.700 --> 01:01:31.290
Josh Moore: I what I can't promise is that it's, it is the final format and you only you know your than done because the rest of the conversation is, what are we going to add to the format. What are all the features we want

457
01:01:32.160 --> 01:01:47.490
Josh Moore: So the agreement that I think we have and that I would like to have with all of you. I think we have it with with a good deal of the community is, you know, we will come up with a specification that everyone feels, you know,

458
01:01:48.630 --> 01:01:59.910
Josh Moore: Represents their immediate needs transform your data, put it online. And at some point in the future, we will need you to upgrade your data on S3, you know, run a transformation.

459
01:02:00.660 --> 01:02:09.390
Josh Moore: That will be version two, and then there'll be a version three. So over the next several years there will be times when you need to take what you've published and update it.

460
01:02:10.440 --> 01:02:27.570
Josh Moore: The benefit you get is, you don't need to run a server, right. So having this format available gets us I think all the boxes you are trying to tick. So I'll walk through this slide, real quick and then I'll shut up because I'm sure everyone has opinions and and or maybe questions, um,

461
01:02:29.130 --> 01:02:38.580
Josh Moore: You know, this is the status quo, as we understand it, which is image comes off the system you have some kind of library and then you can work with. Right. Um,

462
01:02:39.900 --> 01:02:46.140
Josh Moore: If it's simple TIFF, you can work right away. We know all the many formats and then you can have bio formats, um,

463
01:02:47.160 --> 01:02:50.910
Josh Moore: Do the transformation for you ultimately this is time consuming, though.

464
01:02:51.930 --> 01:02:58.980
Josh Moore: So we think what Tsar has enabled is that if you do this conversion that I'm talking about. And you, you have

465
01:03:00.150 --> 01:03:05.190
Josh Moore: Your data and your metadata all in the central than this consistent package.

466
01:03:05.970 --> 01:03:22.320
Josh Moore: That then you can start migrating it to other places. So if you have it locally, you can work with it and you can download it exactly as it is. You don't need any special client, you can just use like an AWS client upload it to the cloud downloaded to the cloud. It's always ice and morphic

467
01:03:23.340 --> 01:03:25.590
Josh Moore: Um, but then

468
01:03:26.670 --> 01:03:38.160
Josh Moore: Any tool that wants to access it, you know, is using its, its low level library. So in, you know, no party you're using the Python libraries and Java will be using. Probably the in five libraries.

469
01:03:38.520 --> 01:03:43.920
Josh Moore: And you're just accessing the data in the cloud as it exists, you know, no servers in place.

470
01:03:44.430 --> 01:03:52.470
Josh Moore: So the next piece and this is just for anyone who's running a Romero that we've worked on is also allowing Romero to export this format.

471
01:03:53.250 --> 01:04:07.260
Josh Moore: So if you run the extra service arm, you could actually point W get or curl or something at Euro marrow and download the Tsar exactly as it exists right we will generate it on the fly. So

472
01:04:07.920 --> 01:04:21.270
Josh Moore: From our side. This is the, you know, if we can make this agreement of the community. We will do everything we can to make this unification sorry this representation. The unified one that we show everywhere, right.

473
01:04:23.880 --> 01:04:24.240
Josh Moore: Okay.

474
01:04:27.450 --> 01:04:28.260
Josh Moore: Any thoughts.

475
01:04:29.250 --> 01:04:33.510
Sebastien Besson: And you're speaking about tooling would just say you said the cost would be the upgrade.

476
01:04:34.020 --> 01:04:40.380
Sebastien Besson: But we will be working on the upgrade tooling any way because we will be testing it ourselves. So certainly

477
01:04:40.800 --> 01:04:52.590
Sebastien Besson: Your. I mean, you're not alone. Right. And that's the biggest cost of development format, probably, is that there will be multiple versions, but I think we will all be in the same boat. So there will be tooling for that.

478
01:04:53.580 --> 01:04:53.970
Josh Moore: Unless

479
01:04:54.000 --> 01:04:57.390
Josh Moore: It's unless a code problem more of a computational problem. So I go ahead

480
01:04:57.810 --> 01:05:07.050
Eric Perlman: Well, and slightly administrative like I like we had with well defined versions, there will be well defined willing to go from version to version of the spec. So I appreciate that.

481
01:05:07.440 --> 01:05:16.500
Eric Perlman: I mean, it sounds like you're getting pretty close to actually. I mean, you've sort of unofficially labeled. This is the first version, given that you have the IPR data accessible in it, whether you like it or not.

482
01:05:17.550 --> 01:05:23.400
Josh Moore: Yeah so version, oh point one certainly exists and we will keep supporting it. Right.

483
01:05:24.750 --> 01:05:32.340
Josh Moore: We're actually at the cusp of needing to go to vo 2.2 so that process is probably going to start ticking along here soon.

484
01:05:34.320 --> 01:05:38.490
Juan Nunez-Iglesias: If I make the other thing that go for you. Oh, sorry.

485
01:05:38.700 --> 01:05:44.070
Guillaume Gay: Thank you. Um, if I understand correctly, would be then easier to

486
01:05:45.270 --> 01:05:56.880
Guillaume Gay: Have a data set in one Romero and move it to a father or a because the use case I'm having are we are planning to have is

487
01:05:58.110 --> 01:06:05.880
Guillaume Gay: We have a marrow one Romero instance pal lab and then publish that data is you know federated

488
01:06:07.320 --> 01:06:16.320
Guillaume Gay: Repository that that's administered by the FBI or things like that. So moving data is turns out to be serious.

489
01:06:18.630 --> 01:06:22.140
Josh Moore: So yeah, so this, this agreement of ISO morph ism.

490
01:06:24.060 --> 01:06:36.750
Josh Moore: I think gets us there. Um, what's missing to do that is a place in the Emmys, our specifications to put everything that's an arrow and we're not there yet, but if you assume

491
01:06:38.730 --> 01:06:44.160
Josh Moore: You're only interested in things that are already in the only means our specification for example like labels.

492
01:06:44.970 --> 01:06:55.740
Josh Moore: Then yes you could you put the service in front of your Romero and you could download the OMB czar and that only means, are you could then put in, you know, you could use anywhere in the same way.

493
01:06:59.310 --> 01:07:15.840
Josh Moore: And as the conversation in the zoom chat is going on you know once there's a czar reader in bio formats, then you could also put that on. These are back into an arrow and it kind of just circles around this graph. I was showing as much as you want it to look just kind of scary. But nevermind.

494
01:07:17.970 --> 01:07:32.190
Juan Nunez-Iglesias: The other point I wanted to make. I think the Tsar binary format is pretty much rock solid the right Josh. So I think a lot of these versions have to do with metadata, which is very small thing to update

495
01:07:34.380 --> 01:07:34.950
Juan Nunez-Iglesias: That's

496
01:07:35.160 --> 01:07:49.890
Josh Moore: True. I'm not. I'm there may be some changes in the Tsar V3 spec that we need to deal with. But in general, what would be really nice if anyone has a solution for this is how in S3 two point

497
01:07:50.880 --> 01:07:59.970
Josh Moore: Multiple buckets to the same underlying data because you're right you know as we go through this process you're updating a little bit of JSON.

498
01:08:00.930 --> 01:08:08.820
Josh Moore: And if you don't have to re copy your, your chunks, then it would be huge. I don't have a solution for it, though, yet. So that's when I'm searching for

499
01:08:11.160 --> 01:08:19.860
Eric Perlman: Us to say, I mean the Tsar is rock solid the legs RB to is missing some things that would be useful. So one of the ones that hit me several times and other people as well.

500
01:08:20.280 --> 01:08:29.910
Eric Perlman: That set that and five has is like the description of how the data are stored. So like for example, if you open up bizarre and you it's stored in directories instead of with dots.

501
01:08:30.150 --> 01:08:41.580
Eric Perlman: Which you want to do on a local file system, you don't care about the cloud. There's nothing in is our metadata to actually tell you that that's the format. It's. And so that's the sort of thing that like will be fixed in this our spec. I think it's going to be. It's in V3 maybe

502
01:08:42.270 --> 01:08:47.460
Eric Perlman: Definitely yes but so that's an example where we actually really would care about the underlines our version.

503
01:08:48.900 --> 01:08:58.350
Eric Perlman: That one is one is really hit me in multiple ways. Because I always want slashes on the local file system and dots on the cloud or I don't care about the cloud, I should say. But I didn't hear about local

504
01:09:00.240 --> 01:09:02.820
Josh Moore: Jason, you were going to chime in a second ago.

505
01:09:03.180 --> 01:09:19.170
Jason Swedlow: Yeah, just repeat what I mentioned in the chat is that while we will obviously do a lot of work to connect tomorrow to these various formats. Nothing in the format should be a mouse specific least that's our intention.

506
01:09:20.460 --> 01:09:26.940
Jason Swedlow: And if we screw that up. We should be told that I'm quite loudly, because that's clearly not what we want to go here, so

507
01:09:30.960 --> 01:09:36.330
Jason Swedlow: There's just, just to say that quite plainly and openly so

508
01:09:36.360 --> 01:09:38.670
Jason Swedlow: Like, in particular, you know, with

509
01:09:40.140 --> 01:09:48.720
Jason Swedlow: Working with one Raphael and others, right, I mean that's that's extremely important to what we're doing here. So I'm just, I'm sorry, I'm doing the policy thing here, and I'll just

510
01:09:49.770 --> 01:09:57.090
Jason Swedlow: If this is kind of the for the avoidance of doubt statement and also just head off back to

511
01:10:01.110 --> 01:10:02.400
Juan Nunez-Iglesias: It. That's great. I'm

512
01:10:03.240 --> 01:10:07.800
David Pinto: Yeah, I was gonna say what I already wrote on the chatroom says, you know,

513
01:10:08.370 --> 01:10:16.080
David Pinto: Coming from the point of, you know, you're already acquired to date on the microscope, you don't have to convert you already have to, you know, save straight into desire for but

514
01:10:16.890 --> 01:10:27.090
David Pinto: You know, basically telling the other researchers that, you know, over time, they have to go back to the raw data and upgraded raw data, you know, and research. I like this petition on the

515
01:10:27.600 --> 01:10:32.430
David Pinto: On the road so that they won't like to, you know, oh wait, there's not the raw data anymore because he already changed it.

516
01:10:32.820 --> 01:10:47.160
David Pinto: So you can like to go and change the mentality. It's okay to change the raw data you're upgrading it you know kind of weaned off. I can see a lot of people going, No, I don't want to touch after position that file is on touch with the

517
01:10:47.160 --> 01:10:47.670
Sacred

518
01:10:49.050 --> 01:11:00.030
Sebastien Besson: But if you don't have a library to read data to producing the first place, which is very much where what we're trying to avoid right essentially the problem, what you're saying is exactly it coming, but

519
01:11:00.510 --> 01:11:11.190
Sebastien Besson: You're so you're assuming that there is a Sacra sent on the flight translation tool which will work for you forever. And that's what we're saying this is not happening.

520
01:11:11.250 --> 01:11:11.880
Yeah.

521
01:11:12.900 --> 01:11:19.680
jean-marie burel: Do you have any question is do you actually convert the raw data or you just convert the wrapping around the rotator

522
01:11:20.040 --> 01:11:33.120
jean-marie burel: Because that's more or less what it is you don't really want to touch the binary, you just change the shell around it. So that's my meta data. My meta data source like pixel size, whatever it is in the file is no longer hidden and

523
01:11:34.590 --> 01:11:39.870
jean-marie burel: And you need a property specific reader for that. This is just coming to light and but the binary itself.

524
01:11:41.040 --> 01:11:47.820
jean-marie burel: Or I kind of this as the same of what you will call the raw data. So I think it's maybe a mental picture, we need to change their

525
01:11:48.780 --> 01:11:49.860
Josh Moore: Marketing question.

526
01:11:49.980 --> 01:11:52.020
David Pinto: You don't have. You don't have to convince me

527
01:11:52.170 --> 01:11:52.980
jean-marie burel: That, oh, I'm just

528
01:11:53.280 --> 01:11:56.130
jean-marie burel: Saying I'm not talking about you, in general, but I think the language to

529
01:11:56.250 --> 01:12:07.530
jean-marie burel: Use BECAUSE I THINK WE THAT IS A we faced in the past have, you know, what are the rendering settings, you know, apply my stuff my channel to read a nice Josh apply to blue and

530
01:12:07.890 --> 01:12:19.260
jean-marie burel: The overriding know we example we will is giving to people is, is having a pair of sunglasses, you know you you look at differently, but you don't change the underlying data. And I think here is the kind of the same similar

531
01:12:19.800 --> 01:12:24.450
jean-marie burel: Approach. If you change the shell but you don't change the raw data himself.

532
01:12:24.720 --> 01:12:25.620
Is to close.

533
01:12:27.000 --> 01:12:28.320
Josh Moore: Because make the data set.

534
01:12:29.280 --> 01:12:36.420
Jean-Karim Heriche: On that just on that point, I from my experience, what actually users called raw data is what is saved.

535
01:12:37.650 --> 01:12:45.120
Jean-Karim Heriche: At the time, or acquisition. So it would help if actually that was the format used by the minute manufacturers.

536
01:12:46.110 --> 01:12:49.290
Josh Moore: Shall we get Stefan to turn on his audio now.

537
01:12:51.930 --> 01:12:53.340
Josh Moore: Can I get a commercial response.

538
01:12:53.430 --> 01:12:54.630
Stephan Wagner-Conrad: Yeah, commercial response.

539
01:12:56.490 --> 01:12:58.200
Stephan Wagner-Conrad: And yes,

540
01:13:01.410 --> 01:13:04.020
Stephan Wagner-Conrad: Yeah, easier said than done, right. So,

541
01:13:06.570 --> 01:13:11.760
Stephan Wagner-Conrad: For some of the for some of the instruments, I can, I can see that this

542
01:13:13.020 --> 01:13:17.610
Stephan Wagner-Conrad: Could be possible. So the question is from from a manufacturing point of view, I think.

543
01:13:20.280 --> 01:13:30.420
Stephan Wagner-Conrad: That's that's there's a stable. Stable specification that we can be sure that you know that that if we implemented.

544
01:13:31.980 --> 01:13:35.550
Stephan Wagner-Conrad: Can can be read elsewhere. And we can

545
01:13:36.690 --> 01:13:37.920
Stephan Wagner-Conrad: Support is

546
01:13:40.050 --> 01:13:41.430
Stephan Wagner-Conrad: With the requirements we need

547
01:13:43.140 --> 01:13:52.440
Stephan Wagner-Conrad: From our acquisition systems. So if the data throughput and expressiveness of the of the metadata model.

548
01:13:53.910 --> 01:13:55.170
Stephan Wagner-Conrad: Is fitting

549
01:13:56.460 --> 01:13:58.770
Stephan Wagner-Conrad: This should be could be possible.

550
01:13:59.880 --> 01:14:03.060
Stephan Wagner-Conrad: Yeah. What, why not, why not so

551
01:14:04.380 --> 01:14:05.610
Stephan Wagner-Conrad: Making no promises, but

552
01:14:07.170 --> 01:14:08.490
Stephan Wagner-Conrad: I can, I can also see that it

553
01:14:08.490 --> 01:14:12.390
Stephan Wagner-Conrad: Makes sense to enable downstream processing pipelines.

554
01:14:13.410 --> 01:14:14.130
Stephan Wagner-Conrad: That's that's

555
01:14:16.380 --> 01:14:18.390
Stephan Wagner-Conrad: Compiled elsewhere, so to say.

556
01:14:21.540 --> 01:14:25.140
Josh Moore: What would make it most likely what can we do to make it most likely

557
01:14:26.130 --> 01:14:26.850
Um,

558
01:14:28.530 --> 01:14:31.050
Stephan Wagner-Conrad: That's a, that's a great question. Um,

559
01:14:32.220 --> 01:14:36.300
Stephan Wagner-Conrad: If it was easy to use to to integrate

560
01:14:39.990 --> 01:14:49.500
Stephan Wagner-Conrad: Yeah and if it was fitting kind of for the for the, for the purpose and so maybe maybe kind of build taking the commercial perspective.

561
01:14:50.610 --> 01:15:02.910
Stephan Wagner-Conrad: For us, there must be kind of kind of some some some advantage. And if it is that we support now writing to the data in a certain format that is required by Institute's

562
01:15:04.170 --> 01:15:07.410
Stephan Wagner-Conrad: Without our long conversions.

563
01:15:07.440 --> 01:15:10.230
Stephan Wagner-Conrad: Of the data and we can support this.

564
01:15:11.310 --> 01:15:32.490
Stephan Wagner-Conrad: Yeah, but this would be kind of a good driver for us and to make it easy. I mean, it needs to be a good specification and libraries that we can use and stable and long time or yeah stable. Stable memory efficient. I don't know all those attributes that you need if you integrate something

565
01:15:34.320 --> 01:15:39.030
Josh Moore: Thank you even want to respond to that add on to that.

566
01:15:43.050 --> 01:15:45.330
Jason Swedlow: I can give the high level response. I mean,

567
01:15:46.740 --> 01:15:50.820
Jason Swedlow: I just like to emphasize this is going to be a long process and

568
01:15:52.140 --> 01:15:59.640
Jason Swedlow: I have to just recognize Stefan has been part of this discussion now for some time and very much appreciate that.

569
01:16:01.320 --> 01:16:04.920
Jason Swedlow: There's, I mean, we talked about this at the only meeting simply

570
01:16:06.420 --> 01:16:18.060
Jason Swedlow: Providing you know the commercial community will need probably.net C sharp libraries, etc. Right. So there's a whole bunch of things.

571
01:16:18.060 --> 01:16:26.550
Jason Swedlow: will not take for granted, which it's not, I could just tell you, I don't know how I'm going to write the grant to get that stuff done so.

572
01:16:29.130 --> 01:16:38.670
Jason Swedlow: Right off the bat, so it's going to be. It's going to be a long process, but I think is Stefan was sort of hinting, there are there are

573
01:16:39.090 --> 01:16:54.780
Jason Swedlow: Pressures, if you will, that we haven't had before. People are depositing data in public resources, you know, the kinds of things that john cream was talking about are now happening are becoming real, you know, there are lots of concrete examples of that.

574
01:16:55.860 --> 01:17:09.870
Jason Swedlow: 510 years ago that wasn't really the case right so there. Things are changing pretty dramatically, and there's a lot of different drivers, then there, then there have been, and all of the hopefully will combined to help

575
01:17:11.130 --> 01:17:25.140
Jason Swedlow: Lots of different members of the community move. I can also tell you in the industrial sector. Sorry for by that I mean things like biotech farmer, etc. There's also a lot of pressure as well. So there's a lot of things happening that

576
01:17:26.160 --> 01:17:37.230
Jason Swedlow: Haven't been in place before but to pretend that we're going to release a Python library. And then all of the commercial manufacturers in the world. They're just suddenly going to adopt that is

577
01:17:39.240 --> 01:17:43.770
Jason Swedlow: As far as least as I know, Stefan might want to comment but is that's probably unrealistic.

578
01:17:45.210 --> 01:17:58.110
Jean-Karim Heriche: But on the other end. The, the old compute model is evolving and and by the particular moving to which PC and clouds. We are moving away from windows.

579
01:17:58.620 --> 01:18:13.440
Jean-Karim Heriche: So at some point we are starting to have this problem, we were, we have those local windows based workstations where and when people need to move the the software to cloud environment. And that's going to work so

580
01:18:14.010 --> 01:18:20.880
Jean-Karim Heriche: My, my feeling is that at some point the manufacturers will also move a little bit in that direction, too.

581
01:18:23.790 --> 01:18:38.550
Jean-Karim Heriche: So I don't think we need to provide these.net also libraries. I think they eventually from the customers pressure there will be some demand for for the the tooling to evolve on the commercial side.

582
01:18:40.020 --> 01:18:45.090
Jason Swedlow: probably agree. JOHN GREEN, but there's probably a five to 10 year transition that we're kind of

583
01:18:45.420 --> 01:18:47.580
Jean-Karim Heriche: Going to be some transition, but

584
01:18:49.200 --> 01:18:54.720
Jean-Karim Heriche: I think we're not going to to move back from that.

585
01:18:57.240 --> 01:19:03.750
Stephan Wagner-Conrad: I just want to add. We also find with c++ native libraries. So, so wrap those. So we're doing this.

586
01:19:05.430 --> 01:19:06.120
Josh Moore: So I so

587
01:19:07.380 --> 01:19:15.090
Josh Moore: I think I mentioned somewhere. I can't remember it's today or in the video. The, the fact that there has been some czar funding specifically

588
01:19:15.630 --> 01:19:29.100
Josh Moore: Part of that money is going to a c++ library for czar. It's from the quants stock group on GitHub. I can paste the links letter later or someone can look it up. It's called X tensor dash czar.

589
01:19:30.480 --> 01:19:35.400
Josh Moore: It's early days, but that may be something where someone who's on the c++ side wants to get involved.

590
01:19:38.670 --> 01:19:51.150
Eric Perlman: Not direct. I mean, or I guess is a tangent. But, uh, Jimmy amount of effort of Google his tensor store library, which is c++ has folds are an invited support as well. And I've been using that recently.

591
01:19:56.970 --> 01:19:58.380
Josh Moore: Thanks to you. Yeah. By all means.

592
01:19:58.830 --> 01:20:00.900
Josh Moore: Everyone can throw those in the hacking D duping if you want

593
01:20:01.860 --> 01:20:05.460
Eric Perlman: Yeah definitely love it because system now and more and more powerful.

594
01:20:06.300 --> 01:20:08.280
Josh Moore: So yeah, for anyone who doesn't know part

595
01:20:08.280 --> 01:20:08.580
Christian Tischer: Of

596
01:20:08.880 --> 01:20:21.420
Josh Moore: The reason that Eric said earlier the Tsar format is so or someone said so solid is that it's so simple, so underneath it's it's actually, it's quite simple to write an implementation, um,

597
01:20:22.470 --> 01:20:27.990
Josh Moore: Now, I don't think there should be 30 or 40 implementations, we should try to keep that low as well but

598
01:20:29.490 --> 01:20:36.870
Josh Moore: At least in terms of the bus factor, you know, worst, worst case some library gets abandoned someone else can take it over so

599
01:20:37.680 --> 01:20:47.700
Eric Perlman: Trevor session. But for example, his work video and supporting last anything else on the client side is fantastic. And I think brings us

600
01:20:49.200 --> 01:20:53.640
Josh Moore: Different places. Yeah, I think a lot of this has really enabled the JavaScript community so

601
01:20:55.500 --> 01:20:55.950
Josh Moore: Cool.

602
01:20:57.090 --> 01:20:58.500
Josh Moore: Any last thoughts otherwise.

603
01:20:59.280 --> 01:21:01.350
Josh Moore: Yeah, maybe. Yeah.

604
01:21:02.670 --> 01:21:07.140
Christian Tischer: I mean, I just better understand now why I haven't been in this discussion, and I think

605
01:21:08.670 --> 01:21:13.860
Christian Tischer: You guys basically as far as that and then not only a patient implementation. So I guess that's why the

606
01:21:14.610 --> 01:21:25.800
Christian Tischer: To say that's more interesting for the typhoon crowd at the moment. Right. But I also appreciate. It's a bit of a chicken and egg problem that you we all have maybe now, right, because I think they all want a specification

607
01:21:27.090 --> 01:21:36.060
Christian Tischer: To do this abstract is, for me at least with tough so much easier if I can try something and say that doesn't work. Let's do it like this. But then you need

608
01:21:36.390 --> 01:21:43.290
Christian Tischer: A library, but maybe you guys don't want to write the library. If you don't know for what yet, right. So, I don't know, I probably we have to hit her a little bit

609
01:21:44.460 --> 01:21:45.360
Christian Tischer: Because I think

610
01:21:46.680 --> 01:21:50.880
Christian Tischer: For me, it's much easier to join the party. I have something to try

611
01:21:52.320 --> 01:22:00.930
Christian Tischer: But that was in my case, Mina Java thing that I can play with. So, so I don't know. Also, I don't know what to recommend, but maybe if it's

612
01:22:02.070 --> 01:22:08.940
Christian Tischer: But my feeling is like getting something in Java and maybe an in c++ that people can play with

613
01:22:10.050 --> 01:22:20.190
Christian Tischer: Maybe rather early and then say guys, this is not able, it will change. But just to get people involved. I mean, I don't know. Yeah. But as I said appreciate. It's not simple.

614
01:22:21.570 --> 01:22:35.040
Josh Moore: Yeah resourcing issue certainly um I mean today. You could take the the czar's that are on S3 download them and access them with the in five star library. So that works.

615
01:22:35.970 --> 01:22:49.110
Josh Moore: Um, the metadata is that's missing is it's quite straightforward. So you know how to do the multi scales and how to set the rendering settings, nothing that you couldn't do if you wanted to, um,

616
01:22:50.250 --> 01:22:57.600
Josh Moore: What is missing and tissue knows this, but this is really for everyone. Since this falls into the cloud section is it's not currently possible to

617
01:22:58.260 --> 01:23:09.930
Josh Moore: Use the Java libraries, the Java in five libraries to talk does are on S3, that doesn't work, and this goes back to what Eric was talking about. There are things that are outside of the specification

618
01:23:11.250 --> 01:23:19.050
Josh Moore: Like the path that he was talking about, but also the S3 storage, um, which makes it very difficult to work.

619
01:23:20.040 --> 01:23:31.920
Josh Moore: To have something that that checks all the features we need in the cloud with czar, blah, blah, blah. So we're working on that. That's really the next step. And what that's the blocker to getting Java working

620
01:23:33.420 --> 01:23:39.090
Josh Moore: So certainly, the more help we can get to do that, the better. But I would hope that happens sometime this year.

621
01:23:44.970 --> 01:23:45.420
Josh Moore: Okay.

622
01:23:46.830 --> 01:23:53.430
Josh Moore: So we're just short of a half hour left. I'd like to get a give everyone a chance to say

623
01:23:54.510 --> 01:23:57.690
Josh Moore: What type of data, they would like to see an oil. These are over time.

624
01:23:59.340 --> 01:24:12.090
Josh Moore: This you're welcome to make this a you know a wish list shopping list, we will probably turn it around, though, and say, you know, can you help right the specification and and maybe open some PRS but it would be good to get a

625
01:24:14.220 --> 01:24:16.440
Josh Moore: Just to capture that list so

626
01:24:18.450 --> 01:24:23.850
Josh Moore: You added. So single molecule imaging right

627
01:24:31.230 --> 01:24:34.800
Josh Moore: Well, I know he did so I'll just he's welcome to say something. He wants

628
01:24:36.270 --> 01:24:45.480
Josh Moore: From our point of view, I mean, I don't have a ton of experience with this my memory of the last time we were investigating this space is that there were multiple transformations to go from

629
01:24:46.650 --> 01:24:49.050
Josh Moore: The single molecule data into an image.

630
01:24:50.310 --> 01:25:02.700
Josh Moore: And so I don't really know where the stands. I am certain we could store the image if someone wants to also store the inputs to create the image, then, that sounds like it would be a spec, a new spec.

631
01:25:03.480 --> 01:25:18.420
Jean-Karim Heriche: And the so you could. So the quite my question, because I'm not familiar with the spec is where whether you you are so that was mentioned before space for story tabular data.

632
01:25:20.250 --> 01:25:27.090
Jean-Karim Heriche: So basically localization microscopy produces a table.

633
01:25:28.230 --> 01:25:36.990
Jean-Karim Heriche: Of coordinates and or some other associated information and that is used to turn it into an image, so I could see

634
01:25:37.380 --> 01:25:51.060
Jean-Karim Heriche: That you could store these with the mentioned rated by people would did the job in the first place. Just like I would see you could save the images with their segmentation mask and

635
01:25:51.840 --> 01:26:03.330
Jean-Karim Heriche: The associated object features that have been derived from them. So basically this is the kind of information that I'd like to have in in those two cases at least

636
01:26:04.560 --> 01:26:13.050
Sebastien Besson: So I was going to say yes, same feeling is that all the action we're talking single molecule localization data, which is really the stable thing.

637
01:26:13.830 --> 01:26:19.920
Sebastien Besson: Immediately when the other completely option is that you could potentially reuse what we discussed

638
01:26:20.310 --> 01:26:27.690
Sebastien Besson: Above with the polygons, because essentially you're storing the set of vertices right. I mean, so it might be possible to kind of

639
01:26:28.110 --> 01:26:40.680
Sebastien Besson: shoe horn. But, and we use these concepts to actually store the initial position and then yes, it's absolutely a transformation from these table representation to an image by the end of the day, which will happen.

640
01:26:41.100 --> 01:27:03.450
Jean-Karim Heriche: Yeah, but so the transformation, you need to store these additional information. So basically you have three main column x, y and Zed typically plus actually know for you have the frame and and then you have some information about the the photon. The some some likelihood measurements of

641
01:27:04.560 --> 01:27:24.630
Jean-Karim Heriche: Things and then there is all kinds of statistical process to generate in the image based on this information. So it's a table with probably seven or eight columns of information that contains so probably if you start with the worst stuff. Then you have the you have the noise. So if

642
01:27:26.280 --> 01:27:45.300
Jean-Karim Heriche: I mean, I would basically stole the says a table of feature table, but it would be a pretty that's being raw data. In the case of localization microscopy and a feature table would be object features in the case of when there is a segmentation mask associated so

643
01:27:46.080 --> 01:27:51.990
Josh Moore: big thumbs up from my side. I mean, my take home would be, we need to work on table so that can up the priority.

644
01:27:52.770 --> 01:27:56.130
Josh Moore: The conversation that's going on in the zoom chat, though, it makes me worry so there's

645
01:27:56.760 --> 01:28:07.710
Josh Moore: There's going to be a lot of back and forth about what, what are the columns in this table. And that's going to have to be someone else driving that. So anyone who wants to take that and run with it, we'd love to have the input.

646
01:28:10.110 --> 01:28:13.860
Josh Moore: Are there any other kind of fundamental data types that are missing.

647
01:28:16.830 --> 01:28:28.800
Josh Moore: So as a reminder for me in G, F. F is a hierarchy of multi dimensional arrays with metadata. Right. That's our starting point and then

648
01:28:30.030 --> 01:28:31.830
Josh Moore: You're saying what you want to turn that into

649
01:28:33.360 --> 01:28:33.630
So,

650
01:28:35.550 --> 01:28:36.000
Jean-Karim Heriche: Go ahead.

651
01:28:40.230 --> 01:28:53.100
Will Moore: Um, I guess one thing that has occurred to me recently is that that everything we've done so far is kind of hard coded five dimensional again very much influenced by a mirror me

652
01:28:54.450 --> 01:29:03.600
Will Moore: And it, which I guess would be good to move to n dimensional sooner rather than later, because already we're getting more and more tools that are assuming

653
01:29:04.380 --> 01:29:23.640
Will Moore: Those five dimensions as they are and not and that would be such a big major breaking change. So, to try and to try and make that jump sooner rather than later would be good. I guess that comes down to different types of file formats as well which is some, some are more than five dimensions.

654
01:29:27.120 --> 01:29:39.330
Jean-Karim Heriche: Yeah, so, so the from the meta data. I think it's a matter of making sure that there is no ambiguity in our the meta data relates to other piece of information inside

655
01:29:41.040 --> 01:29:51.240
Jean-Karim Heriche: The file and then are the. The other point is to define the scope of the meta data, how far, what kind of meta data you want. I mean,

656
01:29:52.080 --> 01:30:04.050
Jean-Karim Heriche: I know depending on your use case. You want one type or another. For example, I know a lot of people care a lot about the microscopy and physics meta data from

657
01:30:04.410 --> 01:30:19.560
Jean-Karim Heriche: From the acquisition, but for my job. I care much more about the experimental meta data. And basically what I can say is the biological entity represented in one channel or the other and on things that that

658
01:30:23.670 --> 01:30:24.960
Josh Moore: Yes, to some degree.

659
01:30:26.340 --> 01:30:28.680
Josh Moore: We're going to have to spread this work around so they're

660
01:30:29.700 --> 01:30:40.080
Josh Moore: On images. See, there was a call for an another call apologies. I think that's planned on intended to happen on the 20th of November around metadata.

661
01:30:41.100 --> 01:30:48.570
Josh Moore: So metadata in the sense that you just spelled it out. Sean cream. So basically the the acquisition meta data, perhaps less about the sample metadata.

662
01:30:49.140 --> 01:30:56.760
Josh Moore: Um, I think from our side, what we would more focus on is is a way to let everyone represent what they want to represent in

663
01:30:57.270 --> 01:31:03.090
Josh Moore: The base structure. So, as our gives us the hierarchy of n dimensional arrays with JSON.

664
01:31:03.960 --> 01:31:12.480
Josh Moore: You know what, we're working through in the specifications are ways of using that. And so will say in the JSON so and so is where you should put something

665
01:31:13.230 --> 01:31:19.830
Josh Moore: And then if someone then on you know one layer higher wants to go. Okay, you've told me where to put something in the JSON. I'm now going to put

666
01:31:21.720 --> 01:31:22.800
Josh Moore: Sample information.

667
01:31:24.570 --> 01:31:30.000
Josh Moore: Then big thumbs up, you know. So I think that's exactly the type of ecosystem, we should we should be developing here.

668
01:31:31.680 --> 01:31:32.790
Josh Moore: Or not quite there yet.

669
01:31:35.490 --> 01:31:38.790
Josh Moore: But that's certainly what I was talking about the the position from

670
01:31:39.540 --> 01:31:47.910
Josh Moore: The Wellcome Trust on metadata for N G, F, F. It's basically someone to spend time thinking about how are we going to represent us in JSON. So basically,

671
01:31:48.210 --> 01:31:58.650
Josh Moore: Minimally what's annoying me XML but really taking that further. Just like we can use our to go from five d to Indy, we should also use this as a step to really have an extensible metadata framework.

672
01:32:05.040 --> 01:32:06.390
Josh Moore: Any other kind of high level.

673
01:32:07.620 --> 01:32:07.980
Josh Moore: List socialist

674
01:32:08.010 --> 01:32:12.300
Josh Moore: I mean, you guys are coming up with far less than I was. Fearing. So this is actually great.

675
01:32:12.690 --> 01:32:20.940
Jean-Karim Heriche: So 111 thing I'm, I'm wondering is that if you have some sort of JSON representation for the meta data.

676
01:32:22.260 --> 01:32:28.410
Jean-Karim Heriche: It's unclear to me how you're going to link it to some of the actual image data.

677
01:32:29.610 --> 01:32:33.330
Jean-Karim Heriche: Or is it going to be a system based on IDs or

678
01:32:34.560 --> 01:32:46.440
Josh Moore: Thinking ideas. Yeah, so I mean so far. What it looks like is when you start up so you start to walk down the Tsar which could also be a HDFS file, you know, so you're starting somewhere.

679
01:32:47.070 --> 01:33:04.230
Josh Moore: You you reach either. So for everyone's benefit in these formats including HTML5 and czar and five. There are two primitives right there's either a group IE folder of stuff or their stuff right and folders can have folders, of course. So you have a hierarchy.

680
01:33:05.910 --> 01:33:12.360
Josh Moore: On each of the folders or on each of the the the arrays, you have metadata. So how we've been

681
01:33:12.870 --> 01:33:24.930
Josh Moore: Building on these are to date, and this is certainly something that's up for discussion as as you walk down that hierarchy you you hit something, you hit a let's take the high content screen for as an example. You hit a folder.

682
01:33:25.380 --> 01:33:37.260
Josh Moore: And you look in the folders metadata and you see it's a at plate, you know, so that the metadata on the folder defines the type of the folder. It says everything that's in this this folder is a plate.

683
01:33:38.310 --> 01:33:44.460
Josh Moore: And then you have the metadata that tells you what to expect inside the folder. You know, I have this many rows this many columns.

684
01:33:45.180 --> 01:33:51.300
Josh Moore: This many acquisitions and then you know how to go down deeper until you get to another folder, which is the image.

685
01:33:52.020 --> 01:33:59.790
Josh Moore: And that, you know, you look in the metadata for that folder. It says, Oh, I'm a multi resolution image and then you know how many resolutions, there are

686
01:34:00.270 --> 01:34:10.680
Josh Moore: And then finally, after this whole process, you get to the arrays. Right. But it's always a case of you're attaching the metadata for a thing to the thing itself.

687
01:34:16.320 --> 01:34:33.240
Will Moore: But we're already not entirely strict about that and the plate. The next level down from the plate is columns and the next level down from that is rows that we store the rows and columns, both on the plate not and and the number of fields so

688
01:34:33.870 --> 01:34:37.830
Josh Moore: Yes. I mean, this is the one under discussion. So if we get into this, you'll have the Olympic

689
01:34:37.860 --> 01:34:49.080
Josh Moore: Team kind of bickering back and forth with each other, which is fine. I mean, this is the fun of what this is, is, you know, trying to come up with these specifications is a process of working through them and if we want to do that. So from my point of view. So just a

690
01:34:50.160 --> 01:35:03.000
Josh Moore: Wills. Right. You know, I don't think the column folder is a thing. It's in there. But what I didn't say is, every folder has to be has to have metadata. Right.

691
01:35:04.560 --> 01:35:05.100
Josh Moore: So,

692
01:35:06.570 --> 01:35:16.020
Josh Moore: It is a balancing act between making it too complicated and no one's going to be able to use it and I'm completely

693
01:35:17.970 --> 01:35:25.290
Josh Moore: Spread out right so it's tricky and we're not sure we have all the answers. That's why it's great to have as much feedback as possible.

694
01:35:33.270 --> 01:35:35.160
Josh Moore: Okay, I see some movement on the next

695
01:35:35.250 --> 01:35:36.690
Josh Moore: Day, go for it.

696
01:35:36.840 --> 01:35:40.080
Juan Nunez-Iglesias: Say, well one thing about this whole thing is

697
01:35:41.400 --> 01:35:48.540
Juan Nunez-Iglesias: For me, I would really like metadata to be has optional as possible because I'm often sort of

698
01:35:49.230 --> 01:35:55.740
Juan Nunez-Iglesias: You know, writing an image from scratch, in a sense, like I don't actually have them intubate. I'm just trying to make a dummy image that's going to work with all this stuff.

699
01:35:56.610 --> 01:36:05.010
Juan Nunez-Iglesias: And so I don't want to have to figure out a million metadata expect fields. I want to figure out like a very minimal set that will make the tools work so

700
01:36:06.570 --> 01:36:07.860
Juan Nunez-Iglesias: Yeah, that would be my thing is

701
01:36:08.970 --> 01:36:10.770
Juan Nunez-Iglesias: Trying to make things as optional as possible.

702
01:36:12.480 --> 01:36:17.790
Will Moore: Yeah, so I guess just coming back to the plate example, some of what we've added there is because

703
01:36:18.630 --> 01:36:34.410
Will Moore: If we put it on S3, he can interrogate what the loan directories are we actually have to list them. Whereas if you're just writing something locally and you could interrogate works in the in the container, then we wouldn't actually need that. I don't know if that's

704
01:36:35.700 --> 01:36:36.090
Will Moore: Useful.

705
01:36:37.770 --> 01:36:38.040
Will Moore: Yeah.

706
01:36:42.600 --> 01:36:43.770
Josh Moore: But yeah, point taken

707
01:36:46.680 --> 01:36:56.640
Juan Nunez-Iglesias: I would say with plates, probably one thing that things like rows and columns seem somewhat specific to a plate geometry as opposed to something

708
01:36:58.110 --> 01:36:58.560
Sebastien Besson: Yeah.

709
01:36:58.740 --> 01:37:00.900
Juan Nunez-Iglesias: I don't know. Maybe someone comes up with a long strip of things.

710
01:37:01.230 --> 01:37:09.120
Sebastien Besson: Yeah, I mean, there was discussion, just, I mean to expand that we started with played because he is is a big concern. I mean, obviously there's a lot of people wanted to do that but

711
01:37:09.540 --> 01:37:15.930
Sebastien Besson: Very quickly, it appears that the plates may just be effectively a subset of generically.

712
01:37:16.320 --> 01:37:24.690
Sebastien Besson: A layout right then and played appears to be very much present and people want to do real computational work and play. So that's where we started.

713
01:37:25.200 --> 01:37:39.780
Sebastien Besson: But think of stitch made registered image multimodal all of these, you know, or I mean, could be seen also as variation of some of the concept. We're trying it out here. So there might be another level.

714
01:37:40.170 --> 01:37:46.740
Sebastien Besson: Of generalization of how you group a buck or not buckets are a collection of images together.

715
01:37:48.090 --> 01:37:59.940
Sebastien Besson: Again, back to what you're saying. I would save this level of audio group images is completely optional. Right. I mean, the low level is I've got an email with some core metadata that allows to describe it.

716
01:38:09.300 --> 01:38:16.110
Josh Moore: Okay, so we're kind of winding down some. We have two or three more topics that maybe we can just touch on quickly.

717
01:38:18.810 --> 01:38:22.530
Josh Moore: And then before we kind of go back to the high level was this useful to everyone.

718
01:38:23.940 --> 01:38:28.770
Josh Moore: Stefan, you mentioned compression and this is just, we need to think about it or

719
01:38:31.950 --> 01:38:37.380
Stephan Wagner-Conrad: Yeah, I actually wanted to just talk a bit about it.

720
01:38:39.090 --> 01:38:41.970
Stephan Wagner-Conrad: So, um, but, yeah, we've been looking into

721
01:38:43.020 --> 01:38:48.120
Stephan Wagner-Conrad: Some advanced compression algorithms and libraries and the last couple of weeks.

722
01:38:49.200 --> 01:38:57.690
Stephan Wagner-Conrad: Namely else at four and then STD I'm yet to find get to understand better

723
01:39:00.510 --> 01:39:06.090
Stephan Wagner-Conrad: Last less a good compression rates for lossless confessing with higher data rates.

724
01:39:07.380 --> 01:39:12.810
Stephan Wagner-Conrad: For yeah for to use this during acquisition. But basically,

725
01:39:15.180 --> 01:39:26.760
Stephan Wagner-Conrad: So, and those, those two are on our kind of shortlist to be investigated further maybe checking out the prototype. And yeah, so I just wanted to

726
01:39:28.590 --> 01:39:29.250
Stephan Wagner-Conrad: Discuss

727
01:39:30.900 --> 01:39:37.470
Stephan Wagner-Conrad: And maybe get your feedback on what you think about those kinds of formats. If you see alternatives.

728
01:39:38.640 --> 01:39:41.460
Stephan Wagner-Conrad: That should be considered so

729
01:39:43.110 --> 01:39:58.230
Stephan Wagner-Conrad: Yeah, and what kind of is there some I don't know consensus on the under number of different compression methods that should be could be available there. And the next generation format the stereo shortly shortlist Yeah something like this.

730
01:40:03.450 --> 01:40:14.730
Josh Moore: So I added a link to czars current list of compression algorithms, adding them on the Tsar side is fairly easy. I know there's a way to add them on the Java side as well.

731
01:40:15.450 --> 01:40:19.710
Josh Moore: So to some extent that goes beyond this call is that you know the

732
01:40:20.670 --> 01:40:36.360
Josh Moore: The makers of the the underlying formats. The specifications that we're talking about. They need to also agree on which compressions. They're going to to support and we keep talking about Java and Python but then you have Java and Python and C and c++ etc. Um,

733
01:40:38.550 --> 01:40:50.100
Josh Moore: I could certainly see us using our cloud. So as a community to basically say in our space in the OEM is our specification. We say compression type whatever is mandatory.

734
01:40:50.670 --> 01:40:59.310
Josh Moore: And then we feed that back to to these other teams and just go, you know, we really want to rely on this always for every implementation.

735
01:41:02.340 --> 01:41:05.670
Josh Moore: I don't have a personal preference. I would love to see someone do

736
01:41:05.880 --> 01:41:08.580
Josh Moore: Benchmarks and gain some knowledge on these but

737
01:41:09.690 --> 01:41:12.780
Josh Moore: It is certainly a place where you have to have a

738
01:41:13.950 --> 01:41:20.130
Josh Moore: Fine dance, along with the specification that will find bouncing between the specifications and the implementation so

739
01:41:21.720 --> 01:41:22.680
Josh Moore: It's going to be interesting.

740
01:41:24.060 --> 01:41:26.490
Josh Moore: But yeah, if you can keep giving us feedback stuff, and that'd be great.

741
01:41:30.150 --> 01:41:41.610
Josh Moore: Um, the copying data conversation has now gone to five levels of indentation. I'm not going to try to read that while we're talking to someone want to summarize. Are you guys happy. Can you just keep going on your own.

742
01:41:43.260 --> 01:41:45.420
Josh Moore: Me. It's great. I'm not complaining whatsoever.

743
01:41:46.830 --> 01:42:02.580
Christian Tischer: No, I mean, I was just, I mean just right, we mean I have a 4.2 terabyte and five thing in an object store and I've been doing. Try to actually copy it from the object store and

744
01:42:02.940 --> 01:42:03.960
Josh Moore: I think I broke your arm.

745
01:42:05.850 --> 01:42:11.550
Christian Tischer: And from what I got is almost not possible. So I don't know.

746
01:42:13.080 --> 01:42:13.470
Josh Moore: I mean,

747
01:42:14.970 --> 01:42:20.700
Josh Moore: So anything that comes down to network issues and performance, yet it ends up being a lot of finger pointing light, you know,

748
01:42:21.270 --> 01:42:39.870
Josh Moore: That may be something if we could have someone in the ambleside talking to us about what went wrong. I mean, in general, a well run S3 server gets pretty good upload and download upload and download rates, even for lots of files but your mileage can vary so

749
01:42:42.240 --> 01:42:52.890
Simon Li: My experience has been, you get a decent chance surveys, but it latency for the initial request comes to be quite slow. Sometimes even with S3, even with Amazon's S3 service.

750
01:42:55.320 --> 01:43:06.810
Eric Perlman: So it's a very slight tangent to that I noticed there's been some threads on the Tsar GitHub about doing range requests. So I mean, someone's using it to like read an HTML5 file, etc.

751
01:43:07.200 --> 01:43:10.800
Eric Perlman: But I know again from this sort of the stuff with the neuro glance or world.

752
01:43:11.610 --> 01:43:21.180
Eric Perlman: Range requests or an interesting solution to that. So you can maybe a bird that might end up being a version of bizarre back end where everything is concatenate it into a small number of files with an index.

753
01:43:21.600 --> 01:43:34.230
Eric Perlman: And again, this idea where oh these are mostly about the metadata on top of czar being a back end, that might be something to look into going forward in most S3 providers and web providers port range requests.

754
01:43:34.800 --> 01:43:43.500
Josh Moore: So for everyone's benefit. There are a couple of different projects Trevor will be here up this afternoon. He's written a couple of these. So

755
01:43:44.550 --> 01:43:52.080
Josh Moore: It may actually end up being a an extension point. So basically what it comes down to as a way to say I've got a biggest file.

756
01:43:53.310 --> 01:44:06.120
Josh Moore: And it can be Tiff and I want to write in some metadata, how to go from the the block, you want to look at to offset and the big file right

757
01:44:06.780 --> 01:44:23.370
Josh Moore: Um, and I know of three or four implementations. It's been done with TIFF, it's been done with HTML5. And so, yeah, that may be the solution if if there's enough work going into standardizing that then that may be the solution, so we'll see.

758
01:44:23.730 --> 01:44:37.680
Eric Perlman: I think it will matter with the multi terabyte data sets that people need to start moving them around and it's not something that's also something you would only do or say you only do for read only data set. So like way conversion of a static data set.

759
01:44:38.400 --> 01:44:41.250
Eric Perlman: But yeah, as Trevor's examples I've been great.

760
01:44:42.120 --> 01:44:42.360
Yeah.

761
01:44:43.980 --> 01:44:49.800
Josh Moore: Okay, so yeah agreed lots of work there. And I can tell you it was the first thing, Paul said when we first discussed this in

762
01:44:51.120 --> 01:45:03.060
Josh Moore: I want one file. You know I want all my images in one file that I can drag and drop on the browser. Sorry. On the Finder. So these are things we need to think about is just the usability of the format.

763
01:45:03.810 --> 01:45:06.990
Eric Perlman: I just want to make it very clear, I'm not proposing that we use zip store.

764
01:45:08.850 --> 01:45:09.750
Josh Moore: Yeah.

765
01:45:12.270 --> 01:45:14.910
Josh Moore: Okay, last point metadata.

766
01:45:15.960 --> 01:45:28.500
Josh Moore: David, you asked, I answered. Dom says that is it is only to read our files. Yeah, so we don't necessarily have a czar writer at the moment. I mean there is, there are methods and only means our pie for writing, um,

767
01:45:29.310 --> 01:45:42.360
Josh Moore: It's a work in progress, but certainly if before someone writes their own code to write like in Python to write one means ours. Just put it in the base library. We're happy to have you know additions or whatever. So

768
01:45:45.810 --> 01:45:46.080
Juan Nunez-Iglesias: I

769
01:45:46.470 --> 01:45:48.420
Juan Nunez-Iglesias: Was working on a writer plugin.

770
01:45:49.620 --> 01:45:50.220
Josh Moore: Who is

771
01:45:51.210 --> 01:45:54.360
Josh Moore: Gaga. Ah, excellent. Perfect, yeah.

772
01:45:55.800 --> 01:45:58.410
Josh Moore: I thought I I seem to remember something like that.

773
01:45:59.910 --> 01:46:00.480
Josh Moore: Okay.

774
01:46:01.650 --> 01:46:03.690
Josh Moore: I still have some closing things I wanted to go through

775
01:46:04.710 --> 01:46:05.730
Josh Moore: Anyone else have

776
01:46:06.960 --> 01:46:10.980
Josh Moore: Core thoughts. They want to get off their chest heads desks.

777
01:46:17.280 --> 01:46:29.310
Josh Moore: Where's our doubt. Excellent. Just the state. I wanted to have everyone in. Okay, so I'm assuming, since everyone was okay with the recording. You're also okay with us publishing the notes somewhere.

778
01:46:30.360 --> 01:46:35.340
Josh Moore: And we'll get these things up and just basically have a record to everyone who couldn't be here could

779
01:46:36.600 --> 01:46:37.800
Josh Moore: Participate in some form.

780
01:46:38.910 --> 01:46:43.200
Josh Moore: If you do have any objections of that form just send an email or get in touch, somehow, some way

781
01:46:45.960 --> 01:46:57.540
Josh Moore: Then it would kind of be a question of, do we want to keep doing this, and if so, first question when I said in the original image SC post roughly once a month.

782
01:46:58.950 --> 01:47:00.240
Josh Moore: One time up from Ron

783
01:47:02.460 --> 01:47:04.860
Josh Moore: Is this useful enough to do it in another month or

784
01:47:06.090 --> 01:47:07.110
Jean-Karim Heriche: I don't know if it's

785
01:47:09.060 --> 01:47:11.100
Jean-Karim Heriche: too frequent. It depends on on

786
01:47:12.600 --> 01:47:28.890
Jean-Karim Heriche: On the level of the discussion for for the level with it here. I think it's probably too frequent but for people who are actually getting their hands dirty in the actual implementations probably once a month is a good frequency

787
01:47:30.720 --> 01:47:41.550
Josh Moore: So from our side on roughly the scale of a month, we would hope to have the HTTPS more hammered out and published to the community. So, you know,

788
01:47:43.320 --> 01:47:51.780
Josh Moore: I could you know you saw or some of you saw I guess wills demo of how it's working right now throughout November will work on

789
01:47:53.040 --> 01:47:59.790
Josh Moore: finalizing that and getting the spec published to image se so a time for comment. So you could also comment at a meeting in a month.

790
01:48:00.180 --> 01:48:09.420
Josh Moore: And will be preparing hopefully the Java work with tissue for images to knowledge. So hopefully we'll have working Java examples to go along with the Python examples.

791
01:48:09.780 --> 01:48:15.840
Josh Moore: So that's from our side. But obviously, if any of you guys are going to do something in the next month you would be welcome to

792
01:48:16.620 --> 01:48:29.760
Josh Moore: Present or ask questions or whatever, you know john cream. For example, if you want to start converting or taking some of our existing data. I mean, some of the data that's up there is yours. So, you know, having your feedback on that would be great. Um,

793
01:48:33.270 --> 01:48:38.370
Josh Moore: Yeah. And maybe we get different people. Next time, so it's not that every you're not signing up for anything but

794
01:48:39.180 --> 01:48:49.800
Josh Moore: Um, I guess, unless someone has another proposal, I would start the process in roughly a month of asking if anyone's interested. And if there's no interest that's fine will delay it

795
01:48:50.820 --> 01:48:55.440
Josh Moore: I assume not much will happen, December but okay. A couple of thumbs up and nodding.

796
01:48:56.670 --> 01:48:57.630
Josh Moore: Topics

797
01:48:58.680 --> 01:49:05.700
Josh Moore: I mentioned the topics from our side. If anyone has one they want to throw out right now, then we can go ahead and plan for that.

798
01:49:11.730 --> 01:49:23.940
Josh Moore: I'm looking at people. No one's saying anything. So, you know, I guess. Be thinking about it, anything any incremental step towards this you know grand vision would be appreciated, so keep an eye out.

799
01:49:24.960 --> 01:49:27.330
Josh Moore: Same thing goes for recording the videos yourself.

800
01:49:29.610 --> 01:49:31.290
Josh Moore: I guess it would be nice to hear feedback.

801
01:49:32.310 --> 01:49:42.990
Josh Moore: I'm under the impression under this current world we live in that, you know, having short little videos that people can share around is actually a really good way to communicate.

802
01:49:43.620 --> 01:49:54.390
Josh Moore: Um, I guess that'll remain to be seen if other of you want to create such videos and share them, but I think it's worked well for us. It worked well me 2020 and

803
01:49:56.010 --> 01:50:02.700
Josh Moore: I like to see more than rather than less. So actually this occurred to me from there is someone who works on the Nepali team Philip

804
01:50:03.300 --> 01:50:12.840
Josh Moore: And he, he just walked through one of his PR. I think it took like 30 minutes and you just kind of went through it. And for me, that was great because I can't make it to mini California meetings.

805
01:50:13.350 --> 01:50:22.650
Josh Moore: So to keep this kind of going, the more you can record for all our possible time zones, the better. So that's mostly a take home.

806
01:50:23.730 --> 01:50:29.310
Josh Moore: And then my biggest question was how annoying. Was it setting this up strictly on image SC

807
01:50:30.630 --> 01:50:36.360
Josh Moore: I got yelled at by my team because they're like we didn't get any of the notifications. We don't know what the hell is going on. Where should we find this. I don't know.

808
01:50:38.280 --> 01:50:45.540
Jean-Karim Heriche: I think it worked well I just kept get all the notifications from older ones will say, me too.

809
01:50:46.530 --> 01:50:48.420
Josh Moore: Okay, so the meteors aren't good good

810
01:50:48.510 --> 01:50:50.820
Jean-Karim Heriche: Yeah. But otherwise, I think that's fine.

811
01:50:53.190 --> 01:50:58.920
Eric Perlman: You also did a great job of cross referencing it on every source. I wouldn't look that good.

812
01:51:00.990 --> 01:51:05.280
Josh Moore: Thank you, Eric. That just means that you and I are looking at the same things though so

813
01:51:06.750 --> 01:51:12.960
Josh Moore: I guess what I'm looking for the people who are not happy. I definitely appreciate it but I felt like I was missing some people

814
01:51:13.440 --> 01:51:14.970
Eric Perlman: Yeah, they're not here. So they don't have

815
01:51:15.930 --> 01:51:20.430
Josh Moore: Yeah survivor bias. Bias. Right, so how do I get the people who aren't here.

816
01:51:24.930 --> 01:51:39.150
Jason Swedlow: That would require broadcasting on, you know, the euro by matching the national by missing list. And so you'll get you'll get your expand out put up a different party, you know, different crews

817
01:51:43.890 --> 01:51:48.210
Josh Moore: Okay, but no one would say they definitely want to do something different, next time.

818
01:51:49.350 --> 01:51:51.660
Josh Moore: From the people who actually got the notifications.

819
01:51:52.950 --> 01:51:57.360
Josh Moore: Nothing different nothing that was, I mean the options would be do everything by email.

820
01:51:58.710 --> 01:52:04.620
Josh Moore: Setup. There's a note on that set up a slack set up a zoo lab, etc. So

821
01:52:05.820 --> 01:52:14.820
Jean-Karim Heriche: I think the image is always best. I don't want to be on yet another chat stuff so

822
01:52:15.840 --> 01:52:19.770
Jean-Karim Heriche: I would otherwise I'll be all over the place.

823
01:52:20.730 --> 01:52:21.480
Josh Moore: So the OH.

824
01:52:21.720 --> 01:52:29.400
Josh Moore: Ok so i i'm happy to the. That's why I tried it, because I'm, I, I like it quite a lot myself. I think one wants to say something in a second.

825
01:52:29.880 --> 01:52:45.360
Josh Moore: The only thing I can do to make mostly my own life simpler. Um, and I guess Vito looking for vetoes on this would be to set up a group so that I don't have to add each of you individually to get notification. I just add all of you in one word, right.

826
01:52:47.670 --> 01:52:51.060
Jean-Karim Heriche: So on. I mean, it's using discourse. Right.

827
01:52:51.510 --> 01:52:53.550
Josh Moore: Right, it would be a discourse group. Yeah.

828
01:52:53.580 --> 01:52:57.480
Jean-Karim Heriche: Yeah, so you can. So the admins after pro of doing that.

829
01:52:57.750 --> 01:53:07.470
Josh Moore: Yeah, so I can do that. And then we would, you know, and we can all add people to it. So that would be it would essentially be a mailing list, but it would be a discourse mailing list right

830
01:53:09.240 --> 01:53:12.750
Josh Moore: General Interest or any fear of that concept.

831
01:53:13.830 --> 01:53:21.960
Josh Moore: Getting some nodding no fee. I'm assuming David's no fear and Anatol is yes approval. So I just didn't like I said I interpret like I want

832
01:53:24.390 --> 01:53:24.960
Josh Moore: Okay.

833
01:53:25.980 --> 01:53:27.540
Anatole Chessel: I think me not to cry.

834
01:53:27.780 --> 01:53:46.680
Anatole Chessel: Ahead to get a crowd it meets also down to people who get the medication to send the mail around cuz I know in my lab people in a first up with not not as a know that cool but may be interested. But it's why I just hope.

835
01:53:47.610 --> 01:53:47.970
Josh Moore: Okay, so

836
01:53:49.650 --> 01:53:54.900
Josh Moore: So I trust all of you to contact others if you think it's pertinent. Yes. Excellent.

837
01:53:58.200 --> 01:53:59.760
Josh Moore: Three minutes for closing statements.

838
01:54:00.810 --> 01:54:01.590
Josh Moore: And I want to say anything.

839
01:54:06.270 --> 01:54:17.880
Jason Swedlow: It can make a soft shoe cost and having now suppose in as a few times. So trying to do this video thing is the short lightning talks is an idea for us.

840
01:54:18.930 --> 01:54:27.930
Jason Swedlow: It's a way to keep the meeting focused on the discussion and also have resources that we can all link to and actually see what each other are doing

841
01:54:29.340 --> 01:54:40.590
Jason Swedlow: We've done it a few times, with all respect, it's really hard to get people to record videos, actually, as it turns out, has been the experience and probably lots of good reasons for that.

842
01:54:41.520 --> 01:54:57.510
Jason Swedlow: However, it's a really great way to broadcast. What we're each doing. And so if it's possible to have, you know, so, you know, so I'm creating you know what is happening in BL. What's the vision look like maybe that's just two slides, but just to talk it through.

843
01:54:59.490 --> 01:55:12.180
Jason Swedlow: You know, Eric, you know, what do you guys do. What are you actually doing. And you know what is the current workflow look like, right. So here I'm you know I'm throwing work at other people. My team knows that's what I that's my job. But anyway.

844
01:55:13.860 --> 01:55:24.510
Jason Swedlow: You know tissue. You know what, what is the status of what you guys are doing. I, I have some understanding through various discussions, but I think in a fairly serious way.

845
01:55:25.020 --> 01:55:33.960
Jason Swedlow: Communicating that at least, you know, as in as up to date form as possible to, for example, this group, but also obviously they

846
01:55:34.620 --> 01:55:47.250
Jason Swedlow: Become useful resources. We're not talking about a 20 or 30 minute talk, we're talking about something that's on the order of five to six minutes. So it's, it's an appeal against Josh, you can slap me down and told me to shut up.

847
01:55:49.980 --> 01:55:51.210
Josh Moore: I'm all on board. I think they're great.

848
01:55:53.670 --> 01:55:54.780
Josh Moore: So yeah, I mean,

849
01:55:57.450 --> 01:56:14.700
Josh Moore: I guess some of you may have seen on images. See, I also tried to come up with a feed like basically i don't i don't think our conversations on Twitter are doing it right. So this is, this is kind of a way to get us a platform where we can have more serious forward momentum beyond Twitter.

850
01:56:16.020 --> 01:56:17.130
Josh Moore: I may be wrong in that, but

851
01:56:17.670 --> 01:56:21.510
Jean-Karim Heriche: No, I, I agree. I think these

852
01:56:22.530 --> 01:56:27.000
Jean-Karim Heriche: platforms like discourse. I actually kind of the best medium.

853
01:56:29.100 --> 01:56:38.130
Jean-Karim Heriche: Because all the others. They are just so instant that you don't have access to, to, to very much of the context of these three

854
01:56:39.210 --> 01:56:43.770
Jean-Karim Heriche: Or it's difficult to to be date and here you kind of get

855
01:56:44.940 --> 01:56:50.310
Jean-Karim Heriche: The kind of mating structure without having actually to have emails.

856
01:56:53.460 --> 01:56:54.540
Josh Moore: All very good feedback.

857
01:56:56.520 --> 01:56:57.660
Josh Moore: All right, we're at time

858
01:56:59.070 --> 01:57:01.380
Josh Moore: I don't mind sitting around drinking a cup of coffee, but

859
01:57:02.550 --> 01:57:07.290
Josh Moore: Let everyone go. Thanks for coming. Everyone, you'll get the notifications for the next ones.

860
01:57:08.220 --> 01:57:09.390
Jean-Karim Heriche: Thanks for organizing

861
01:57:09.750 --> 01:57:10.680
Josh Moore: Yeah, anytime well

862
01:57:10.770 --> 01:57:11.700
Once a month, I guess.

863
01:57:14.700 --> 01:57:15.030
Jean-Karim Heriche: All right.

864
01:57:15.180 --> 01:57:15.600
Josh Moore: Take care.

865
01:57:15.660 --> 01:57:16.890
Jean-Karim Heriche: Bye. Okay, bye.

866
01:57:18.210 --> 01:57:19.080
Christian Tischer: Thank you very much.

867
01:57:20.280 --> 01:57:20.910
Juan Nunez-Iglesias: Thanks. Bye.