 Hi, it's afternoon. Can you hear me there? Maybe, I don't know if I don't need. Thank you for checking in. So I'm very glad to present today. These two initiatives are more community-driven initiatives to aim into my computer vision and about my data science more accessible. I'm myself, I'm a research fellow at the Anduin Institute working for the data science and humanity, and I'm honored to present with others who are part of this project, the insight vision project that was injected at the tour. So, yeah. Hi, everybody. I'm Alden. I am a senior researcher for research applications at the Terran, the tools, practices, and systems program. And it's been great to see so many of you on your posters talking about data and interoperability and when we use that will really focus on the tools, practices and systems. One of the tools for this is called SciVision. So it's a toolkit for scientific image analysis. Basically, we want to recognize the fact that a lot of different problems that use imaging and computer vision to analyze those images are facing similar challenges, you know, they might be from absolutely different domains. So I'm actually a neurobiologist by training. That's my background. But we've got people like we've seen today looking at tree crowns, you know, different mapping and astronomy even we've had some cool collaborations through SciVision. And we just make a generic platform that will allow people to discover and share methods and computer vision as well as data. So they can find what rather than inventing a new a new version of something that's already been done, go and discover something that exists that could work for your data. So the mission of SciVision to solve real world challenges and democratize computer vision to support interdisciplinary and international researchers. And basically the idea is that we have, you know, people developing algorithms developing models using computer vision, and we have data owners and we want to bring them together. So we have this catalog, where people can, people can upload either models themselves so we have, we have several pre trained models in there. But we're trying to collect a broader array of different models and have them in the catalog so people can see how they work, and what kind of data they might, they might be useful for, as well as a catalog of data sets so that if you have data you can necessarily know how to either way, if you have data and you want to find a model to apply to it you could try looking at the catalog for models, see if anything might be effective for your data but you could also upload data that someone else could then use and try to try to test out a new model or create a new model using that. So I'm in the lead, I'll have to speak a little more about the specifics. Okay, there are some things like, as I said, one of the users that we identified so far are developers, so we are aiming to enable these kind of users, oh, users are not developers to distribute their tools. So we have the front end, if you visit now submission is the front end, and now you easily can add your model and you can have the model into two kind of options. We've been loadable in submission, we have IPI in Python, and just provided the decal repository, we at the moment try to work with these two kind of inputs from the developers community. And as well, for data, those are the data producers, and as well they can submit their data to this catalog that we call so far. And in this case you can see at the versatility of at the moment the current image that we have, we have a bio-imaging examples and models of plant life science and as well something about environmental science. And it will power these data providers and discover this lattice computational algorithms and apply them to these different images. And finally, we are also aiming to providing a bridge between different data scales and formats. It's not related at all, but we have as well a project start where you can find different projects that being like pairing these data sets and models that are available in the catalog, and you can also well run some notebooks or demonstrators that we are also trying to maintain as a tutoring, but as well, anyone can submit their projects and who would like to have more visibility on their projects using computer vision models. So at the tutoring, we saw how this idea start with SCOTS that was here and it was one of the PIs and we have another PIs working in life science and working more in agriculture. But in the case of the environmental sustainability with SCOTS, we start with these projects and this is kind of the challenge that you can find in related computer vision. We saw this one of all three ground, but there is also a lot of like a super resolution on the diffusion in satellite images and something about very important here at the petition that the survey is this is where preparation and tracking. But you see, sorry, in the plant life, in the plant science or in the culture there are images that have all kind of challenges you can find images that are 3D. And that's why my colleague Evie, that is part of the submission team, she had this kind of challenge to create computer vision models to try to do this kind of seed pods from this kind of CT scans. My colleague and her team is working as well with this plant phenotyping images as well, very common among environmental scientists is using this satellite images to do some classification of cementation. And very interesting is about this bio-imaging. When I start working with this diverse group of postdoc, I know that they are working as well with very noisy images and if you see this and you say what they are extracting for these images. And I don't know if you're familiar with protein analysis, but it's very complex, very high dimensional data set that they play with, but in some how the algorithm that they are using for tracking this analysis, these images can be transferred to our own images in environmental science. It's very interesting. For instance, they are doing a lot of graph neural networks for mapping and separating different proteins from these images that we don't understand, but this kind of algorithm my colleague are working on this. This is a very successful example, it's a vision that why this transfer landing between domains is very relevant at the Turing. There is a research that was developed for analysis of historical maps and they wanted me to use, they developed a kind of like deep learning like deep learning models for doing some patch-based mapping of highways and certain roads in front of historical UK maps and what my colleague did was to try the same model, but in order to try at the depth, leaves and flowers and she used this same patch-based classification algorithm that was initially developed for that specific purpose and she transferred successfully to help this particular specific kind of application. So this is the only thing that we are aiming to facilitate and foster around the community to try to transfer this different knowledge among the different domains and try to be a platform where people can explore this very easily. We don't take a long time of the priority areas, but just to let you know, apart from having your contribution model and data sets, we are working very strong in certain areas. One of that is the core features. I would like to say that there is a hard work here because sometimes models they develop with certain dependencies and so they work in terms of flow set and version and sometimes they work with biters, if you are familiar with the libraries. And handling lists and having those servable to run with that image is very challenging. So we are working on that, how to handle these dependencies conflicts. In terms of the automation, we are working in facilitating how users can start creating their own data set and login data sets using template repositories, because in order to be loadable in SyVision, you need to reformat your code to be readable for SyVision. We are working on that. In terms of the web interface, something that is not available now is that you go to a data set, SyVision is, we will have a kind of future maybe related with the iPad is going to suggest which model you can run this data set with. We are working on something that we are working, we are working to improving the model cards. And in terms of community engagement, I guess it's very important here. And that's why I started this. And then that is part of the DPS team, they are helping us to grow as a community because there is a huge opportunity to here to connect different domains and as well to create a hub for best practices how to share computer vision models and how to share data sets. We are trying to lead these aspects at least through the SyVision. And finally, we would like to support use cases and that's why here we are presenting this in the environmental community and trying to see if you have any computer related vision, as well as data set model, you can share it. These are the kind of tasks that we are focusing in the next three years, SyVision is funded for the next three years, and we are aiming to have more samples of our super resolution, and samples of our object tracking, and we are aiming to support use cases like that. And we are checking the identification that is common across domains and very much based research challenges. So that's the vision. We have a newsletter that everyone is welcome. If you want to learn more, you're welcome to scan this QR code. Clearly, we have a newsletter, we are aiming to facilitate this conversation across domains. Okay, I guess I haven't finished. So now we move to EDS Book. EDS Book is another community driven initiative. It's inspired by the Turing Way, maybe you're not familiar with the Turing Way, but Turing Way is a community as well. And therefore that is aiming to generate more knowledge and try to have more collaborative ethical and reproducible data science, and it's a global community, but essentially I started this project, because I found opportunity, how we as environmental can communicate how our staff could be data set, research, or pilots through Jupiter Nobus, maybe may ask, who are familiar with Jupiter Nobus? Are you, maybe not all people are familiar with Jupiter Nobus, so it's good to us. We also have interactive computing, where you have a narrative, and you can add code, and you can run this in an interactive manner. You don't need to run like from beginning to end in one run, but you can run each of the cells in somehow. This is the main concept, so essentially we are exploiting this format to build this community that is aiming to highlight what we are doing in environmental data science. So essentially you can say here, we have these levels of use cases, and essentially we are maximizing this format to produce new ideas, and in the team that you will see later, but I mean to highlight reproducible, discoverable, and charitable environmental data science. So, sorry, I don't know what to do with this. So our mission is to educate and leverage good scientific software and data management practices among environmental scientists to review fair executable nobots. Fair is a common line running, but essentially it's the way forward that now people are thinking about group practices in terms of data sharing, so what I mean that these executable nobots as a software are as well findable, accessible, and interoperable, and reusable. And our mission is that us as environmental scientists were collaborative to demonstrate and communicate science through fair executable nobots, and we have gained significant skill to publish in the workspace publication system, what I'm making emphasis here because I think the future of digital is like that, like people share the narrative but you click on the figure and you can see what is the code that you generate this figure. So this is part of the format and I guess one of the format that is been prototyping for that in the future is the Jupiter file format, and there is a data initiative in the American Geoscience Unit that is nobots out and they are meant to have a prototype this system. And this boot is training in researches in that way, in that in that live vision of having this as a primary format for publication. So this is the gallery that we have. So far you go to this boot is being nobota being brought to by the community, and essentially we have a gallery of these fair nobots, and we have in the title we have some touch and say this story about this topic we're trying to have these budgets checking if it is something that is useful in the future, but most important here is about having this batch about the review, so people can check what was the first draft of the nobots and they can see how with the comments of the community who review them on nobots, these nobots have the final version so they mean is to improve and having high quality nobots. This is very common and it's not what we are trying to check that they are running a project how much and so I won't go in the technicalities of this. And this is a summary of this command community like noboots so essentially someone like me that was doing my PhD four years ago, I went to a noboot and I found, I can already know but I don't know which is the data provider that are making a clear statement of who are the original co-base of all things, all things we can improve like through community, and that's why we have reviewers, we have people who kind of did that, and we are finishing I don't know. Sure are you okay. Okay, okay. And at the end, very important is that we are I mean that is noboots you can reuse for all the purposes and that's why in this situation we have different boots of different colors. So that's something that you can achieve when you have high quality nobots. I won't go in details but we are knowing me to be a publishing journal where I mean just two people get familiar how you can publish open source software and we are taking some stages. I don't know if you're familiar with the journal of open source software and they have these stages where you have the review review so this kind of as well as the training process that you learn. So it's something that we are making proof in the platform, maybe the future because at the moment everything is in Github and there are people that even they don't know how to do comments in Github and there is a long process. So, I guess it's something that we've been learning through this initiative. Key achievements, we have some guidance, we have templates about how to start your first notebook, we have at the moment the noboots and these noboots are reproducible in binder. We are trying to have this fair research object where people know can know all the metadata of the noboots so people can know which is the contact or software requirements to run this in your local computer or how you can run this in binder, all this kind of information we are trying to track through a fair platform that is suited to have fair research objects. And finally, we have some community meetings that we are no longer running. And basically because it's very demanding to maintain a community like this we are in so how partnering with the during way and we are in so how we are, we are hosting this community meetings now in the during way collaboration cafes. And presently, and thanks to this technology that infrastructure that we have provided we run this with the climate informatics, a company that was a Cambridge in it was in April. We run this repository challenge, and we use infrastructure here, or they get good to order to save people okay we have this paper that was published in it is your not that is it from Cambridge University Press, we say to the party once please can you get the same resource, but in this case try to make your reports using interactive you put the noboots, and that was a kind of the approach that we use it the challenge. So we very innovative in run the challenge because they were you they use the stupid and almost so if you want to see the report on how they were successful you can run. How they generate their possibility that so it's something that is very interesting about the challenge. And something that I would like to highlight is that it's not what's that is still working progress but we are going to release then soon in the in this book gallery. So we have three things at the end completed the challenge that was very interesting. As well something about nice about running this with you put the noboots if you got many contributions, and in particular because we are doing this per review process we have people who do the reviewer or who will do the judging we have some speakers and we have some different contributions that you can see here we did this with the Cambridge C scale provide resources and similar in our way that as well help us to organize this. And finally interest or priority areas about this initiative is something that is this platform aims to be a platform where at least during is aiming to publish like demonstrate certain research of doing research and collaborators and I think that they just will be a central platform where you will see to the researchers and analysis in environmental science. But for that we need to prove some confusion but not that you cannot many annotations so we are going to increase information in the noble format life, which is a software requirements inside the metadata of the noble which is a kind of information so there are many things going on there, and in terms of the floor name we would like to have the user experience and the gallery does. And this is very important about the multilayer noboots is something that I would like to do maybe the noboots to experiment is like in the noboots you can see the scientific outcome but if I am decision maker, I would like to only read this narrative that is specifically for decision maker have kind of filter I say I would like to bring only that narrative that is essential for decision maker say I would that's what I think we made by what we like your noboots. And finally, we are supporting a not only Python noboots we are supporting by a noboots in Julia are in JavaScript, and we would like to host more activities co-op more activities with research network like we did with the climate science that was very valuable experience for the community and to improve the infrastructure of this resource, and you can follow us this year's staff in our different channels of communication that we have. Thank you. Thank you. Can I ask a question, please. So regarding side vision. What are some advantages of creating a platform for ecology or nature ML models, instead of using a general platform like hugging face. Well, I would say that there's, please. Yeah, we have, we actually have overlap so you know platform like hugging face. And maybe the same model, ideally, you know these models can be shared in both places right because they're open. We actually have a couple of models inside vision that are also from from hugging face. So we would say the advantage is that side vision specifically about computer vision that image data, hugging face would be a broader a broader kind of platform, but also very valuable. John still busy. So is anybody else got a question. When compiling images from different sources inside vision. Do you have any issues with sharing, e.g. use usage attribution. The only woman and data that we have is data that we know that they have the license to share. And of course we have some problems like private private data that one colleague is using for platform typing this data can be shared. With the side vision API, you can run in your own computer and cheating using internally to demonstrate the capabilities with internal catalogs working only in their own server so we can support both kind of that like what it is that one that the attribution is declared that can be shared, but those one that is private, of course they need to use internally with the side vision API in the one to run the API. Okay, thank you very much. Okay, and it's time for Rachel.