Advanced search

Message boards : News : [wildlife] paper accepted at ICCS 2015

Author Message
Travis Desell
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 16 Jan 12
Posts: 1813
Combined Credit: 23,514,257
DNA@Home: 293,563
SubsetSum@Home: 349,212
Wildlife@Home: 22,871,482
Wildlife@Home Watched: 212,926s
Wildlife@Home Events: 51
Climate Tweets: 21
Images Observed: 774

              
Message 5030 - Posted: 10 Mar 2015, 16:51:31 UTC
Last modified: 10 Mar 2015, 17:00:37 UTC

I'm really happy to announce that we've recently had a paper published on Wildlife@home in the 15th International Conference on Computational Science (http://www.iccs-meeting.org/iccs2015/). We compared user accuracy to expert accuracy for both the old interface and the new interface, and results are pretty great -- with the new interface some categories (like bird presence) jumped from around 50% (basically a guess) to 70-80%!

If you'd like to give the paper a read I'm making a pre-print version available for you here: http://volunteer.cs.und.edu/csg/wildlife/publications/2015_iccs_wildlife.pdf. Let me know if you have any questions!

Thanks again for all your support, feedback and participation in the project! On another note, I'm putting the finishing touches on a new Wildlife@Home application, and next week is spring break, so expect test workunits out this week and weekend for that.

Profile JumpinJohnny
Avatar
Send message
Joined: 24 Sep 13
Posts: 237
Combined Credit: 10,275,610
DNA@Home: 192,548
SubsetSum@Home: 201,740
Wildlife@Home: 9,881,323
Wildlife@Home Watched: 55,997,833s
Wildlife@Home Events: 15,584
Climate Tweets: 333
Images Observed: 351

              
Message 5031 - Posted: 10 Mar 2015, 18:20:18 UTC - in response to Message 5030.
Last modified: 10 Mar 2015, 18:22:45 UTC

Congratulations!
Thanks for sharing the paper. I hope it generates interest and opens subject dialogue in the 2015 Conference.
As for the accuracy rates increasing for bird presence, I think it was mostly a result of several changes to the new interface.
1. The longer videos available allow the viewer to identify bird movement. The grouse sit very still for long periods of time and purposefully camouflage themselves while incubating. Depending on camera view and obstructions, it can be almost impossible to determine over shorter intervals whether there is actually a bird on the nest.
At night it can be assumed they are on the nest but sometimes certainty can only be gained by seeing them there at dusk and daybreak. Even relying on eye reflections as a cue to prove presence in the dark is made easier by having the longer video lengths.

2. With the new choice of viewing sequential videos of the same nest it can be determined from a previous video if bird is present or not at the beginning of the next time segment. That is much easier than guessing what was happening on the nest when viewing random short segments on random nests as with the old interface. (I still like being able to pick a "random" next video when I get bored of a certain nest.)

3. Being able to change Event times at review and alter existing Event names also allows some increase in bird presence accuracy. (I think that creating an Event at review is still NOT a good Idea and would be detrimental to the science and results.)
____________

Travis Desell
Volunteer moderator
Project administrator
Project developer
Project scientist
Send message
Joined: 16 Jan 12
Posts: 1813
Combined Credit: 23,514,257
DNA@Home: 293,563
SubsetSum@Home: 349,212
Wildlife@Home: 22,871,482
Wildlife@Home Watched: 212,926s
Wildlife@Home Events: 51
Climate Tweets: 21
Images Observed: 774

              
Message 5033 - Posted: 20 Mar 2015, 16:18:42 UTC - in response to Message 5031.


1. The longer videos available allow the viewer to identify bird movement. The grouse sit very still for long periods of time and purposefully camouflage themselves while incubating. Depending on camera view and obstructions, it can be almost impossible to determine over shorter intervals whether there is actually a bird on the nest.
At night it can be assumed they are on the nest but sometimes certainty can only be gained by seeing them there at dusk and daybreak. Even relying on eye reflections as a cue to prove presence in the dark is made easier by having the longer video lengths.


I agree, this was probably the major part.


2. With the new choice of viewing sequential videos of the same nest it can be determined from a previous video if bird is present or not at the beginning of the next time segment. That is much easier than guessing what was happening on the nest when viewing random short segments on random nests as with the old interface. (I still like being able to pick a "random" next video when I get bored of a certain nest.)


This is probably something we could look into more -- how often users viewed sequential vs. random videos.


3. Being able to change Event times at review and alter existing Event names also allows some increase in bird presence accuracy. (I think that creating an Event at review is still NOT a good Idea and would be detrimental to the science and results.)


Yeah, I'm with you on not allowing events to be added/removed at review. Then a user could simply mark up videos and then re-do everything afterwards.

On another note, Kyle has also been working on some new background subtraction methods which we're getting fairly good performance with. So we're hoping to test out an algorithm which should be able to pre-screen videos for events of interest, which hopefully will further help things.


Post to thread

Message boards : News : [wildlife] paper accepted at ICCS 2015