January 2018
Intermediate to advanced
310 pages
7h 48m
English
The motion of objects in videos has very good information about the actions performed in the video. The motion of objects can be quantified by optical flow. Simonyan and Zisserman (http://papers.nips.cc/paper/5353-two-stream-convolutional-networks-for-action-recognition-in-videos.pdf) proposed a method for action recognition that uses two streams from images and optical flow.
Optical flow measures the motion by quantifying the relative movement between the observer and scene. A detailed lecture on optical flow can be found at https://www.youtube.com/watch?v=5VyLAH8BhF8. The optical flow can be obtained by running the following command:
p1, st, err = cv2.calcOpticalFlowPyrLK(old_gray, frame_gray, ...
Read now
Unlock full access