Wildlife@Home: Some more initial results

[RSS]

Here's some figures of the initial data analysis. In the first set of charts, we used the motion detection algorithm and compared video segments containing any event of interest (predator presence, bird leave, bird return, nest defense, nest success), to those without any active event. http://people.cs.und.edu/~tdesell/sharptailed_grouse__active_1.png http://people.cs.und.edu/~tdesell/sharptailed_grouse__active_0.png While not great, there is definitely a noticeable difference in having higher likelihood's with active events. It does give us motivation though, that we're making progress and will be able to come to a good solution to this problem! The second sets of figures are for detecting the presence of piping plover: http://people.cs.und.edu/~tdesell/piping_plover__plover_all_1_1.png http://people.cs.und.edu/~tdesell/piping_plover__plover_all_1_0.png Similar to the motion detection, there was a notable improvement, but not yet enough for it to be relied upon. However, we still have many things we want to try in the feature detection arena, so we expect to be able to dramatically improve on this. cheers, and thanks for all your hard work, time, and support! --Travis

More... (http://volunteer.cs.und.edu/wildlife/forum_thread.php?id=126)