The following links direct to 5 different demo videos.
Badger 1
Badger 2
Fox 1
Fox 2
Fox 3

In the above attached videos, you will see video frames like the following.

Subplot a: a input frame in a video.
Subplot b: the average variation of the activated pixels in the current frame. A threshold is set to 20. A frame whose average variation above 20 is sent for classification.
Subplot c: the cleared version of the activated pixels, with optical flow indicating of the motion estimations.
Subplot d: if a frame is activated, it will be classified by the trained CNN framework, the classification result will be printed on the top right corner.

The recongnition can be achieved by a self-designed CNN framework (CNN-1) or a transfer learning framework based on AlexNet (CNN-2)