 Yeah, so I guess that my presentation is a bit of an outlier compared to the other presentations that we have seen, which were mostly about observers in developed countries. As I was looking at these presentations, it slowly sank into me that the work that I've been doing is actually also with observers, but recruited from the crews of small scale fishing vessels. And I hear you guys thinking, well, you cannot make an official and an observer, but perhaps at the end of this presentation I can show that under some conditions it actually is possible. So I've been working with the crews of about a total 600 Indonesian snapper tuna fishing vessels. We've equipped them with cameras and we've asked them to take pictures of the catch. That is what it boils down to in a nutshell. So this image or this slide already tells half of the story. Here you see a fisherman taking an image of a couple fish put on a measuring board. We have been doing this program between 2015 and actually up till now. We've covered the entire Indonesia Exclusive Economic Zone. We've worked with small boats, but also with some larger boats of about 100 GT. At a global scale, I guess that these larger boats are still, that's only medium size, but for Indonesia they are quite small. So one of the specifics, one of the characteristics of the fisheries in Indonesia is that most of it is small scale or even nano scale, you could say, very small boats, hundreds of thousands of them operating often from very remote landing sites. So that makes it very difficult to even deploy port and numerators, let alone professional observers. So when we were designing our monitoring program, I noticed that my technicians were communicating with the fishers through images, mostly through WhatsApp actually. And over time we decided to capitalize on that idea and to work with these fishers, or at least some of them as observers on their fishing vessels. And we developed a routine that is completely fisher friendly, it is form free and it also doesn't require any filling in of paper forms or electronic forms. It also has its limitations of course, and I will tell you a bit more about these later. So the purpose of the data collection that we were doing is to get length based data for length based stock assessment, perhaps I don't really need to go into the specifics of it, but it boils down to a length frequency distribution of an exploited fish stock, having a different shape from a length frequency distribution of a fish stock that is not exploited or less exploited. So just from the shape of a length frequency distribution, you can say something about the exploitation status of the fish stock. And as for many of Indonesia's fisheries, we don't know anything about exploitation status. That is a good place to start. So we were interested in catch composition and we wanted to work with these fishers with cameras. So each of the fishers we work with, we gave them a pocket camera. We use pocket cameras rather than hand phones because pocket cameras work way faster. They are way easier to operate than a hand phone camera. On the hand phone camera, there's just too many other apps and especially if the screen gets wet, it becomes difficult to operate. And for people who are working on board and who want to work quickly, it is really important that they have a piece of equipment that they can just switch on, that they immediately is ready for taking pictures. Also in the way that we work, the fisher hands over the data at the end of the trip to a technician. And that is also a bit easier with memory cards that are found in pocket cameras. Obviously, we included the size reference in each of the frames of the pictures. So on the top right here, you see a measuring board and that measuring board is really only as a size reference. It's not used by the fishers as a measuring board. We also used other kinds of tokens and references depending on the type of fishery. One challenge remained the total catch. For some of the fisheries, especially for example Snapper, that is a one by one fishery. We could be reasonably sure that the fishers took pictures of most of the fish that they caught. But for other types of fisheries, pursaners and pollen liners, that was not always possible. So in those cases, we asked the fishers to also take an image of the sales receipts that they already have. In addition to get some more social economic data, we asked them to take an image of the receipts for the supplies that they bought. So we are basically working with the information systems that the fishers already have, basic as they are. We just take it one step further. We also equipped the boats and you see that these are really small boats with tracking devices. They became quite important in our program. We used spot trace. Spot trace is not even a real tracking device as such. It is actually marketed as an anti theft device. But they are real time and they are battery operated. They have an on and off button button, which means that the fishers actually can switch them off. But we felt that they still yield very useful data for us. And here you see the fishing positions of drop liners and long liners in the Arafura Sea. You actually do see that here on this image that the fishers even report their locations when they know that they are fishing in areas where they are not supposed to fish. They need the waters of Timor-Leste and even to some extent also the waters in Australian waters. Indonesian fishers are quite known to fish the line. Our system does require quite a bit of a human support system. We use one technician per 10 or 20 vessels. And then we also have one data quality control supervisor per 5 to 10 technicians. All depends a bit on the type of fishery and the size of boats that are in each unit. We also developed an online portal for data suppression for data quality control and for reporting. So some of the key attributes of our monitoring program. First of all, this is longitudinal monitoring. So we are following a relatively small set of vessels, so about 600 vessels through time. It's not cross sectional. Many other monitoring programs, like for example port sampling, basically go to a beach or a landing site or a harbor. And then you take observations on each boat that happens to land at that day. For us it was a bit different. We followed the 600 vessels through time. Obviously, the monitoring program does require some training. Fishers need to learn how to take good images and what to take images off in the first place. Especially in the beginning we have issues with fishers taking pictures only of the fish that they like most. And that of course doesn't give a good representation of the total catch. We found that the role of the technician is quite critical as a liaison with the fisher. The technician usually does a debrief with the fisher more often so in the beginning of his engagement than towards the end. Because after a couple of trips the fisher pretty much knows what to do and data quality control is already quite good. We found that an incentive is necessary, so a simple fee works best. So the fishers we work with do receive a small fee for the work that they are doing. We also think that this way of working is probably not suitable to roll it out to all fishing vessels in Indonesia. But it works very well with a small percentage of the fleet based on their representation. So if you're looking at a fishery that has gill netters and say handliners, then you recruit a couple of each of those fishers. And in that way you get data from both handliners and gill netters. So some of the benefits of the program that we worked on, we got high resolution data. The timestamp of the images is really useful, not only for geo referencing, but also to get an idea of when exactly the pictures were taken. So if you see that an official takes images only on the last day, you may not know that there is something wrong with the way that he is working. Obviously the data are verifiable, I don't need to explain more about that. But in Indonesia this is particularly important because we have such biodiverse fisheries. So the snapper fishery that we've been working on has more than 50 common species. As I've shown earlier, even though it is a voluntary system, it does generate data that is relevant to compliance. And finally, we also found that this network that we developed of collaborating fishers is quite useful. So it is a good sounding board for developing ideas, for example new ideas on management. We also commissioned a couple of studies on artificial intelligence. The results were quite good. One of the methods that we developed is, you see highlighted here, the machine that you see on the picture on the deck of a fishing vessel is designed to take standardized image of fish. So it's not working with those snapshots that we have been collecting. It aims to get really fully standardized images. It worked really well. So the accuracy overall on the data set of 44 species was about 95%. We used 2,500 training images, then we trialed it with 766 images. And here you see an example of three critical moiety snappers snappers that are almost identical. I mean, you do see differences here on this image, but even a seasoned observer would have difficulties telling them apart. The AI system, however, had no problems whatsoever telling the difference between these species. Sorry for these three species, 100% accuracy rate. My final slide has some recommendations and insights. First of all, even with this fairly simple way of using cameras, onboard procedures for data collection still depends a lot on the type of gear and on the type of fishery. So you need to tweak it if you're going to work in a different fishery. It makes quite a difference whether you're working on a one-by-one fishery on snapper or a percane fishery. We found that the estimation of total catch does remain a bit of a challenge. So we had to rely on pictures of the sales receipts, for example, for the percane fisheries. Also, if you would roll out this idea to an entire fishery, I think that you would need to rely on AI to speed up the data processing process. We also found that this way of working is actually quite robust to the effects of COVID. At least in Indonesia, observer programs basically came to a standstill, but the work that we have been doing more or less continued, simply because the fishery still continued. There were some delays in getting the chips with the data, but overall, our time series will not show a data gap. So with our voluntary and image-based data collection system, we collected data from over 25,000 fishing tricks, and that proved to be a credible basis for length-based talk assessments. We have transferred these data to the Ministry of Marine Affairs and Fisheries, who is now using these data to compile harvest strategies for various snapper and grouper species we've been working with. So that's it for me. Thank you for your attention. Of course, if there are questions, I'd gladly take them.