 Okay, well, let's get to our last presentation now. And this presentation by Amel Zoubi of the National Institute of Science and Technology of the Sea in Tunisia, providing insight into the need for playing AI to meet the rising demands of seafood sustainability. Welcome again, Amel. Good morning. I am very glad to participate in this event today and share with you my knowledge. I am Amel Zoubi, an engineer consultant in sustainable aquaculture using artificial and intelligent tools. Today, I will present to you briefly some examples about the use of AI tools in fish farming in Tunisia. Our work context, as you know, the demand for fish and other seafood is raising is essential. We need this need sustainably. For offshore cage, we need to produce fish in a way that protects our sea while feeding growing population. The way we manage our sea will determine our future. We believe that by making fish farming better, we can protect our strength in that future for all. By making fish farming better, we can do it. My experience as fish farming technical manager, combined with my environmental background, has convinced me that farmers aspiring to be truly successful in aquaculture will need to be visibly sustainable. AI and data are the most credible way to share the evidence of this. We are working with many projects with the goal of helping Tunisian farmers reduce their environmental footprint. This will help the industry save billion kilograms of CO2 and innovation is the key. Feeding represents the biggest cost for fish farmers and the waste feed had the largest environmental impact. So optimization in this area will always be better and give better profitability. Feeding strategy use site conditions and biomass measures. So to reduce waste feed, feeding protocols depend directly on average weight of fish and appetite, as well as site conditions. So doing good sampling and correct sampling can help to follow correctly the growth. To follow the growth of the cage, the traditional sampling methods is often used sampling methods are different. A traditional sampling done with a net on the surface is less representative than that one done with a camera in the middle of the cage. Taking high quality digital stereo image of this live fish while swimming in the cage. So it's less representative using the traditional way. The principle is stereoscopic video image measures the high and the length of each fish while in algorithm accurately calculate the live weight of the fish. Intelligent software can easily provide detailed biomass report within a accurate size distribution graphic with a lot of measures and statistical parameters. With camera likes camera by Cass HD or camera air max algorithm, or many others, we have an exacter biomass and subsequently a correct feed, which allow us to define a maximum ration, which will be exactly assimilated by the fish and as a result, reduce the feed losses and impact. We can mainly see difference between the minimum weight and the variation. Sorry, we can mainly see difference between the minimum weight and the variation city in the second sampling with the traditional sampling deep net. We take a lot of small fish that they always remain on the surface and the age of the cage, but which to not represent the state of the entire population of the kids way. We can to with this algorithm and intelligent software task have the air max recommended feeding table. For example, for this case, should trash in average weight of around 370 gram at the start of August 2021 or sure respecting free trade equal to air max and take in consideration the optimal generic condition of the sites. The potential of AI in aquaculture does not simply end with local farm economists. The system can provide the potential to power data driving inside into aiding the environmental and sustainability drives. The potential of AI in aquaculture does not simply end with local farm economists. The system can provide the potential to power data driving in safe into aiding the environmental and sustainably drives on this sector. The AI algorithm can measures the poor feeding that contributes to damage floor bed in provides real time and safe before this opens in the future. It's our aim to increase this sustainably efforts. Thank you for listening and I hope you all have a productive day. And if you have any question, please do not hesitate to ask me. Thank you. Thank you very much, Amel. I think there's in my mind some very clear questions here of an overlap between fisheries and aquaculture. We know that in the Pacific, for instance, they're dropping divers into the into the per se net to have a look at the type of fish that are there to decide whether they're going to pull those fish or not. And I think these kind of systems that you've put together with the stereo cameras, I'm just interested in how quickly the data can come back to is there any onboard analysis of these machines so that I mean, obviously in an aquaculture net, you can afford to wait a few days to get the results. But what's the timeline between maybe doing the recording and then getting a result back for your manager? Because I can see this kind of opportunity being very handy, for example, for people working in per seing vessels who want to drop into a net that's closing and decide whether they want to pull those fish or release them and equally for if there was any particular critical species in there instigating a mechanism to release those before they brought on brought the net on board. Thank you. Thank you very much. Yes, we have a return back quickly. Let's say quickly around one hour after doing this video. But the problem is analyzing this video take some hours for doing all the analyze that we need. The other problems are the time that we need for many cages because as you know, we don't have only one cage to analyze each time. For farmers, we have at least 10 cage at minimum and the Portuguese in farmers, for example. In many farms, we have more than 60 cages. So taking a lot of time analyzing each one one by one, we take a lot of time and we can improve a lot in this area. Matt, would you have a question for Amel? Yes, I was just wondering what's it like integrating these systems with the farmers? What's the process and what can you do to improve the farmers opinion regarding using the AI? We have to improve. Normally, for me, we have to improve in three levels. These levels are the time in analyzing the data after taking the video and the time taking the video too. The second level is that for the moment, this kind of software take image in good condition. And we don't have all the time weather good conditions in the site. And the real conditions, we are a little bit far from the reality to help pretty the farmers. The second one is the prices of this because in big farms, it's not possible to use only one software to take a lot of videos for all the cage at the same time and at the same day. So I think that in these three levels, we need to improve a lot. And for sure, it can help a lot for developing sustainable agriculture for sure. I know we've got Amel. I know we've got another question for you. And this is from Anton Ellenbrook. Thanks, Anton. Sorry, hello. Thank you for the very nice presentation. Thank you for it, Joe. Yes, so Tunisia is not so far from Rome. You can see it. For sure. But this software is often developed for other species than the one you grow in Tunisia. But we also work with farms, for instance, in Greece or in Spain, where they may be already using this. And we also work with the feeding industry because the first slide was very good, where you say that most of the costs are in feed. So our collaboration with the feeding industry here seems very obvious. For sure, we are working with others. But to be honest, we start working with the industry of fish feed. They propose to use this software to be sure about the profitability of the project and the result about this sketch to help us to improve the production and to improve the use of our feed. For example, we are working a lot with Italy because the feed is coming from 80% from Italy to Tunisian farmers. And they are doing a good job for a search and for helping us to improve the use of this kind of software. Thank you very much, Amel. You're welcome. We've heard from our presenters today and we're going to have a little bit of a round-table discussion. I just wanted to go through a little bit about what we've seen today. So for example, we started off with Millie as well as others talking about equity. And we've also spoken about the range of opportunities that are out there for doing this type of work. I mean, there is almost no end to when we start opening the box where we're going with this opportunity. We've spoken about how to build communities and how the United States has been very forward-thinking through NOAA to be able to build this community, to come together and start to celebrate and help each other out in the kinds of ways that they're working. And we've seen the cross-fertilization of different models that have been developed and such like. We've also seen the use of competitions to try and spur along the events. And we've also heard about explainable AI, which is going to be a magnificent opportunity. I don't know exactly how that works for deep learning, but for machine learning, if we can start to pick up what is actually driving this, it allows for a much better iterative improvement. And we've just started to get an idea that it's not so hard to do if we can get the right people to share the right images and get the right backup to get those images tagged. But we've started to realize that this whole procedure is allowing us to bring in more opportunities, more data sets together, because machines can work in very quick time. And it allows us to shorten the time between collecting information and getting the outputs. And I think this is a magnificent opportunity because everyone who works in science knows that sometimes when you bring a paper out, you're really publishing something that you did three, four years ago. And this is nice, but in reality, if we're really going to grab the advances that are coming in this type of field, which are coming so quickly both through technology and through the opportunities, it's going to be great if we can bring these stories alongside the fishermen who really see value in what we're doing. So in this session, we are going to ask the community of presenters to maybe throw in a little bit of their feeling. And then we're going to take you through to a little bit of a simulation about what type of themes came to the top of your mind. So if anyone from the panel would like to just give a bit of an idea of some of the messages they took from today or some of the surprises they took from today, or even some of the things that, you know, we might have missed out because there's obviously lots of areas that we think maybe should have been mentioned. Please just go ahead and speak and I'll pin you to the open panel. It's Anthony from Kitware. I'm happy to start things off here. I think across the talks this morning, we saw a wide range of maturity in the use of AI. Some people are really using it a lot and have been for years. Others are really just getting into it. And in forums and conferences like this that we've been involved with at Kitware, we often see that where there is this huge range of capability in AI by a particular marine science group. And these meetings really serve to bring those new people in and expose them to what's possible and the tools that are available, which is great. And then this is how new people come to the Ami community, how they get into working with Microsoft or wherever else is helping out for their altruistic aims. So I think that this workshop seems to be doing really well in that regard. And I think that these meetings that might be their greatest purpose is to make sure that it is to bring in lots more people who to see what's possible. And I hope the folks who we're hearing in the audience will stay engaged and chat with folks who have these capabilities and try things out because there is an active in Virgin and community and it can be really intimidating to see what other people have done already and how far behind you might be. But I think it's easy to try these things where people not to be intimidated and jump in with both of you. I think Scott and Christian bringing a story a bit further along the value chain. Scott specifically, telling stories that are reaching the consumer. And that's a nice way to bring a much bigger audience to this discussion about how we produce our food and how we wanna produce our food. So Scott, what was your take home after listening to some of the presentations? Well, thank you so much. And I just got so much out of all of your presentations. It's been really, really incredible to link up with you. I think what I took away from this is actually the AI has captured imaginations across the world, obviously. And while it's an incredible tool for research, I also wanna kind of underscore that I also from this presentation have seen it's just an incredible tool for science communication as well. I think that the more that we can utilize these tools and these types of events to really showcase the work that's being done, the more you're likely to be able to captivate the imaginations of the public that you're trying to communicate with and that you're trying to reach in order to effectuate the meaningful change in conservation and science that we see. So obviously as a technical tool, what you're doing is extraordinarily valuable. But I also wanna just underscore that it has immense value outside of just the scientific community and into just the public's imagination. So again, thanks for the opportunity to participate. Yeah, I remember if you're really as old as me, you recognize that in the early days when you went to your boss and said, I want a computer, they would go, no, no, we just don't have budget for that. But now everyone's got computers. In fact, they got more than one computer and no one questions the fact that that's a piece of kit that is basic to your business if you're in our business. And I'm just wondering if the stories that we tell are really about capturing people's imagination so that there is, because to be honest, it's money that makes these things happen. Ingenuity is there throughout the world. But when you put money with ingenuity and with keen people, really things start to happen. And this is where, you know, theming up the story for what needs to be done where and what's been left out of these developments which will help to bring the whole together is something that's very important for us. So I don't know if anyone's got any examples there of how the stories changed the way that their departments have looked at the questions that they're dealing with. I mean, Mel spoke about how the feed operators have played a role to help her industry work on these questions. I mean, what's coming through the door that's making it useful? Threatened species, for example, is another one. So for me, for example, I want to share with you the idea that avocachers to try to be helpful and to be sustainable for the environment and using the software help a lot. And you can't believe how much we can help this industry to develop the activity with a lot of sustainable efforts. And it's so important that many companies develop machine learning algorithms because our work can be faster and efficient more and we can obtain a lot of good results helping the industry of aquaculture too. So I am very, very happy today to participate in this event. Thank you for all the team that's organized this work. And I think that the future is AI and machine learning algorithms. Thank you. Thank you, Amel. I've got to pose a question to Matt. Matt, with Kitware, you've got such an incredible range of work on your system. How are you going to partition that? For example, if we're going to put together proposals to bring the international community, we have to theme that story because giving a very complex story to funding bodies is very hard to work with. What's your feeling about the key themes that need the efforts at an international scale? Obviously, small groups have all their own needs, but at an international scale, what are the key themes in your mind? I guess my comments there would be that some approaches are very general. So there are lots of commonalities across the board in what people are doing, things like trawling pipelines that might be able to be standardized a bit more. So some approaches are very wide-ranging in that they can be applied to a wide range of problems and you might not need to reproduce the wheel at every level. But then you have the opposite side of that where people are also doing very specialized things, like invoking thermal cameras as opposed to optical and then triggering optical detections and running those in certain circumstances. Different types of specialized sensors, like going beyond stereo, but fusing stereo acoustics for measurement and some of these other problems might require very specialized solutions. So you really have the duality where you have some problems that I think can be solved pretty generally but then others that are more specialized. I think a few people point this out but the algorithms are imperfect yet. Sure things are advancing very quickly and I expect they'll continue to over the next five years at a rapid pace. But things still aren't perfect and when you have multi-million dollar industries getting things there and boosting your accuracy will definitely be important over the next few years. To tell on that, I think that the higher level of international picture or take this to funding agencies, UN, et cetera is that AI can be used really effectively by scientists without knowing anything about AI. And that's pretty new over the last year or two or three. And that's Vyami's mission is to enable non-scientists to use deep learning and take advantage of all of that without knowing or caring how it really works without doing any programming. So often you need to know how AI worked but you had to be a programmer to make it work. And now with Vyami and other tools, you don't have to have that. You label some data, you use interfaces, you try stuff out, it doesn't work, you label some more data, you try other things and you don't need to be a computer scientist to make this stuff work for you on these specific problems. And I think that's been a real turning point for Vyami and for the community like this at large and other scientific communities that AI can now be leveraged and adapted in the field by scientists. You don't have to have some PhD in the back room solving your problem and then rolling it out to you. Different problem, you got to go back to that PhD again. You don't have to do that anymore. Now you can get the tool and you don't need that expert help. So that enables AI to be used really widely. Okay, well thank you everyone. I just wanted to allow you to the fact that one of the things we have put in the past to try and develop a collaborative project around some of the work that we've been doing with the beta projects. But these have been small apps around threatened species mainly sharks and rays. And I think there's an opportunity now to look more broadly. And I know there's an EU Horizon Infra 2021 call, for example, that has to be written by September this year. And its focus for this big investment is better use of imaging data. So it speaks very much to the work that we're doing. What I'd like to do is add something to the chat now. And it speaks very much to what Matt was just saying. You know, what are the key questions even if they small or big? And if you don't mind diving over to this, this is just a virtual whiteboard where all you need to do is put your name in and you should be given access. And I'd love people to just add what they think are the big questions. And to add anything to this virtual whiteboard, you just double click on a free space on the themes, for example. After you've doubled space, you add some text and then you click out. And you can see I've just done that, which leaves a message. And what I'd like you to do is just add a theme which you think is the kind of theme that you think should be in such an application. The kind of theme which tells a story to funding bodies but also allows it to capture a group of work which might be useful for collaboration on an international level. And then if you have something very fine scale, just add it to the task. So all you really need to do is double click on a blue spot, type in there, click out and it will leave your message behind. And it's just a way of putting ideas up and you can see there's other ideas there. And we will use this page to help us to structure what we're trying to consider to try and get more funding for this kind of opportunity for people to collaborate. And Anton and Matt, would you like to speak at all to this discussion? I was expecting Matt to go first. Ah, okay, sorry. I was just answering a question in the Q and A. Yeah, I think there's actually really important question that's been posed in the Q and A that I'd like to highlight. There's two actually. And one is about data standardization. Did you see that, Anton? I saw it, yes. Yeah, one something I didn't hear much about is about how to make your data AI ready standard best practices, et cetera. It'd be nice to hear some discussion on this. And I think an FAO is positioning in terms of data collection. That's quite an important point. The other one posed by the same person actually is a key to our development. And Noah is the partnership we're building with the industry, Google, Microsoft, Nvidia, AWS, et cetera. How can other countries take advantage of all of this? Can FAO take this responsibility to bring a platform to enable sharing knowledge and opportunities to partners, et cetera? So in the form of a bridging function. And I think that's a real key point and one of the reasons why we've brought everybody together really isn't it, Tim? Anton, yeah. Maybe also Hassan can say a few things. So he asked already, so how do you prepare your data to make them AI ready? I think on things as standards and best practices. I think that is really what this conference was about. So how do we start to develop these standards for an international community? So not only for a universal research center in Europe or the US, but how do we make this then accessible for also for people in, for instance, around the Mediterranean? I think the question that Hassan poses here has maybe to do also with metadata, not only on the image itself or on the streaming media that you get, but also on who will be able to access it and what will be the attribution to the people that brought the data to the platform. So if you think you can work with a big fishing community, what is the incentive for these fishers to contribute the data? I think that is a needed discussion that FVO will have to have and also one of the reasons we organize it. And then on partnerships, yes, FVO is in a partnership already and a lot of partnerships also with private industry. And I think also they are the idea of a partnership with the UN organization is to make sure that other people know about these initiatives and find a way to also connect to maybe private industry, but through a UN body, so we can help them to take the first steps and to feel more secure in taking the first steps. If you are a country with a, every good fisheries monitoring or a control program or if you have an environmental protection agency that wants to know what the benefits are, I think it's difficult for a lot of countries to step directly to a big private industry and ask them for help or assistance. So these countries, I think they first want to knock on the FVO's door. And so if we have these partnerships, and we do have a few, we can help these countries to really reach out and to learn about first, learn about what is possible with AI, but then also maybe help them to take the first steps in AI-driven analysis of environmental fisheries, social economic problems. Yeah, I think that's a very important point that you bring up there, Anton, because in reality, FVO is not really an organization by itself. It's actually just an amalgamation of the world's governments, instructing a small body to help those governments orientate themselves around questions which need international collaboration. So points that Anton was speaking about there regarding coming together and making some type of guidance for countries on what type of, not necessarily controls or rules, but what kind of standards would they promote? And getting those kinds of relationships ironed out so that we do get that sharing happening with people feeling safe in the room and also for the less advantage to not be left out. So that's gonna be a very important question for us going forward. And there's some places where the kind of international arm maybe doesn't need to be so focused. There's gonna be very, very small use cases which are worked on in private that maybe don't need to be shared. And so there's that mix of things that people have to talk around and there's things that potentially can tell stories or allow transferring outside of this kind of international collaboration.