Dear colleagues,
i’am preparing a (finally commercial) tool to automatically assess behavior using some sophisticated image analysis techniques in conjunction with machine learning.
Have a look at some demo results here: http://oo9.eu (edit 2015/03/01: server UP again)
I am looking for further real world test cases and image/video sequences to qualify animal behavior analysis in particular. Image sequences should be registered from a steady point and shall include time points (best in filename: UTCtime.jpg). If you have some study ongoing and you're willing to give the software a shot, please contact me.
Feedback/discussion on the demo are welcome here, too.
Andreas
http://oo9.eu:60009
Hi - I'm not sure if this would be of interest, but I am preparing to submit a paper which will include ~70 raw behavioural video files available for download and analysis. Can keep you updated as to when this is available if it might be relevant?
Hi Rosie,
sounds great - please notify me, if you published it.
Greets
Andreas
I'm interested, and would like to keep in touch.
At the moment my main problem is manipulating videos on an "old" [six years!] computer, but I may get round that soonish.
Best wishes
John T ([email protected])
Recent behaviour paper, with movies: Biol J Linn Soc, first issue for 2015
Hi I do not have video, but large number of encouter histories of gulls (like 001100101.. lenght about 200, 1=seen 0=not seen in population analysis) which include behaviour from different subgroups. Is it possible to imput that type of data for sorting by behaviour. Do you know such software?
Dear Risto,
the software is about assisting in behavior classification from video/still images that tends to take hours to month from the researcher if unassisted. And it is intended for image like matrices only.
Your data might be a time series of observations ? (1-seen/0-non seen) of individuals that can be grouped to defined subpopulations/subgroups ? If so, what is your analysis goal? I would start simply by plotting time series and aggregating the subgroups to see if individuals or subgroups cluster over time. You might do cluster analysis using R then, don't you?
Thanks, for answer. I am bad in R. Yes, they are yearly time series in days or grouped to 7 day periods. I have done preliminary analysis like in Euring Dunedin and USA (http://www.juvaste.fi/risto/PostISOFuusi.pdf ). OK, I shall test cluster analysis, but I have been thinking that the sorting could be done by artificial intelligence (e.g. Self-organising Map, SOM),
Andreas
I have a lot of video footage of brown bears. But they might not be steady enough to meet your needs, given that the camcorder was handheld.
Stephen,
brown bear video spotting - huh, that sounds very exciting. But unfortunately will not work with the software. Needs closed environment to work. I guess, it will take another 10 years until we will see the first software / processing able to cope with the variability of videos taken in nature.
Anyway - thanks for the offer.
Hi Andreas,
The paper I mentioned was published yesterday, with full open field videos for both studies: http://f1000research.com/articles/4-109/v1
Hi Rosie,
great - thanks a lot for this training stuff!
With such excellent video recordings (excellent orientation points, symmetry, good contrast rat vs. underground) most of the tasks - center zone entries , time in center zone, total distance made - are almost trivial from a computational point of view.
I will concentrate on the "rearing-up" frequency and see if a maschine learning approach might have spared you the time of counting this out manually.
Thanks a lot!
Andreas
- Badges
- Science topic