Motion Detection/Estimation examples in OpenVINO, MediaSDK, OpenCL
Hi, I am trying to get a sense of the Motion Detection and Estimation examples in the above 3 software. The MotionEstimation code in MediaSDK and OpenCL are same and apparently uses a build in accelerator called Video Motion Estimation. The OpenVINO motion detection code uses OpenVX graphs. The outputs of the 2 are different also - in the MediaSDK/OpenCL example the output is an array of motion vectors , in case of OpenVINO, these are bounding boxes.
Would appreciate if someone can throw some light of the differences in input (YUV/RGB/MP4), performance ( msec per fame ), output ( array of vectors, bounding boxes ), intended application (high level CV, vs low level video compression ).