okay so it's rather small but anyway so I will talk to you thank you for your

sorry oh yeah so if thank you for your attendance and staying so late today of course you will be there tomorrow so there's no problem I will talk to you about Genvid which is a software we developed at the University of Liège and I'm a biologist I'm Jean Thien and Laurent is engineer and we developed this

because we need to know how our animal behave in their cage and we can't stay there during the whole day and night so obviously we were not the first one to

think about something to measure the activity of animals and the first one first device that comes in mind is the running wheel so you have a wheel over there and your mice of your rat is just there running the problem is that

what are you doing when your mouse or rat is outside the running wheel you can put rats over there for 24 hours a day and the second problem is that you measure only the motor activity and not the general activity if the rat is just moving the head you can't see it you can you can see it but you can't

measure it I think quite nice that were published when we thought about our solution in an open access publication was using a radar and with a complicated device and but everything was free so that's a good point but

do you really want such a complicated hardware a third method is just video tracking so you put the rat in the cage and you follow him that's a good idea to follow the rats but usually it's costly because there is no free implementation and the second point it's impractical because the software

what the software does is it takes the animal it takes the center of gravity only and then it puts some red points all over the animal so that well your animal is there you know but what it measures is only the center of

gravity so again it's motor activity and not general activity so a final solution is markers like people do in virtual reality systems but really rats don't like these things they just show them and it's over so what we want is

the quantification of oops sorry the quantification oops okay okay the keyboard is better no it's blocked I can go so the the goal is to develop

general monitoring system okay it's there without any marker and using commercially available CCTV and modular free software why commercially available

CCTV because you know we are in a university and we don't have extensive research budgets so roughly the principle we use is the frame difference analysis and if you take I will try it once if you take a rat at one time at

one time and you take a picture of it and then at the second time you take another picture of the rat and here it's every 1 25th of seconds you can compare the two images and you compute the number of pixels that are

different here in green so the setup is the following you have your rat you have a camera you have a computer that analyze the images and then you have two displays one with the animals and if well at the first row you can see

that more clearly then if I move my fingers you have some green pixels showing and after that you can analyze the output with general graph like you want so here are two pictures of our system implemented in a real laboratory so the main thing for a biologist is that the rats are

well the wellness and here it's okay because well they are alone in the cage but at least they have their friends around so for a rat which is a social animal it's very important so if people want details here they are you

can use with our system any camera here you need infrared if you want to use 24 hours of recording because when it's dark or night you need to have something on the screen and for example we use a camera from China which costs

60 euro on eBay and the travel costs are included you can use any video for Linux enabled frame grabber in well we use a frame grabber because we have a camera but here on the laptop you have a webcam so you don't need any

frame grabber at least if the camera the webcam is supported but by video for Linux it will work my webcam is not infrared so it will not work during the night we use an ATI rage 120 now it costs let's say 5 euros and you can

use any computer in Linux running Linux we use a p2 that cost we checked this afternoon 20 35 euro on eBay and on the heart so everything it's 100 euro on the software side you use GNU Linux we use a rather old

version of red hats version 7.3 of course the video for Linux API we have no other dependencies and we have an output in ASCII so you can throw your data anywhere even in a file if you want and then you can visualize your data with GD cilab python imaging library I know name your favorite

software so of course on the scientific basis we did a third theoretical validation and we demonstrate that our system is precise sensitive reproducible with rodents with we compare it with an Atti watch

which is the system for human in measuring their activity and it's stable over time so if you want later on we can go into details about that and if you want to check everything it's published in an open access journal on

the web so and finally let's see what's happening with the rats so the rats has a different circadian rhythm than our hours and the rats are merely quite not I won't say asleep during the day during the light phase of the 24

hours and more active and very active during the dark phase so during our night and so during the day when usually people are around in the lab during their experiments the rat is not moving so this picture can be taken

taken by anything your webcam my webcam Genvid anything but what is interesting is that during the night the rat is moving and here you can see all the green pixels around the rats showing that here it didn't move anything with

the legs so the running wheel would not have kept or detected these movements just the head with the ears over here and the tail and a little bit over here so that's why it's for me very interesting so after that you take

the output and you plug them it one in what we call actigram so it's the amount of movement versus the time and here each block is one day so one two three four five etc and the block the dark block above are the night or dark

phase and day or light phase and everything start at midnight so as you can see the rat is more active during the dark phase than the light phase and this rat was quite interesting because all the time around I don't know ten

eleven we just jump out and did something and then everything else so it's interesting because all these were collected automatically so no intervention the rat is just there food is available water is available and no human intervention so after that we can say okay this what was

complying with the treatment or not compliant or did well or did not well depending of on your experiment so in conclusion we can say that Genvid is sensitive reproducible stable over time more accurate than an active which and we proved it with rodents so if you have some pets and all you can test

them and we can come and we will publish another paper with with them so the one interesting thing is that it's non-invasive and one last thing I just discovered quite a few minutes before the talk is that it's very light

the archive is only 52 kilobytes so it can be set up nearly everywhere so what is more important for us is that it uses commodity hardware and it's free as is in free beer and it's free as is free as in free speech so I want to

thank a lot of people in my lab but I will I want also to thank you for your attention and if you want here are the URL to get the archive the paper but more importantly if you type Genvid in Google you will find them so thank you very much there any question yes as such the system cannot

detect behavior so I mean in the output you can't see at one point sleep or something like that but we developed this system because we used regular

sleep system for rats EEG electrons fallogram and it was very difficult for us to do it in an automated way and we don't have any a lot of space on our hard disk so it was impossible for us to store 24 hours of sleep or not

sleep but we don't know so we store everything so with this system you can say below a certain threshold okay maybe the rat is asleep or not but just record the sleep in case and then after that review the sleep and so as you can see the rat is not not not moving all the time so it's quite

moving very often so it reduces also the amount of electrons program that the scientist has to review but your question is very interesting because it's one of the next step we want to do is that to categorize a certain

pattern of what is what moved if the rat only moved ahead it can means it's only drinking eating or maybe just nothing so it also means it's not sleeping it's not moving so we can categorize certain behavior and that's

what we are kind of working on yes yes and breathing and more importantly everything depends on the quality of your camera so here the camera was very good because we inherited from a very successful lab but we use some

camera with a lot of noise even if there is nothing in front of the camera so there in the certain process you have to define a certain threshold below I don't know 20 pixel in the image I don't want to see anything so after that above 20 pixel yes okay let me show me the data

and let's see but you have the raw data so if you don't know what your camera how your camera will behave or certainly your camera goes wrong you can say okay let's see the data you have to your data and at one point you

say no this is not good you have too many too much noise at one point then you can reduce at one point so you can act before the recording and after no problem yeah so that's a good point that the question was can we use

multiple camera from different angles for example to see all the movements and if you remember all the images are taken from the largest side of the cage

because more of the movements are done in that plane if I can say but it could be interesting to have two cameras and since well the system is not using much of the processor it couldn't be difficult to put the second camera for

the processor and have the two logs at the same time and compare them or merge them into a summary yes so I'd like to say yeah so I just so I always like to end with an open

source talk that's a bit different but believe it or not three talks ago I actually saw a nice crawling out of the ceiling running here in front and going

back and I wish I could have taken a picture if only I fed your system I would have it okay so thank you for coming to the lightning talks can I just shortly ask if you would take any trash that's around you throw it there that will save us some work and then we'll see you tomorrow thank you very much