Win10, Anconda Env, DeepLabCut 2.0.4.1, CPU

Feature request: speed improvement when analyzing videos with CPU (and GPU)

Solution1:
Select FPS when analyzing video (e.g. recorded with 120 FPS, but only 60 FPS needed)

cap = cv2.VideoCapture()
cap.set(cv2.cv.CV_CAP_PROP_FPS, 60)

Solution2:
Create roi around animal when analyzing video (detection of points in first frame) and looking for matches within the roi instead of entire cropped frame. The roi in next frame is expected to be near the last detected points.
(E.g. mouse in one half of the arena cannot jump to the other half within one frame; therefore the video has to be analyzed only in a fraction of the frame.

Cheers ;)