I’m trying to use the optical flow SDK 3.0 to compute the flow vector to replace the motion vector from NVENC by Jetson Xavier.
Since I acquired the flow data from CNvOpticalFlowOCV.cpp
I found the size of array seems not the global one and I have no clue how to arrange the vector back to the original picture.
How can I know the placement rule of the array?
And If it is based on the ROI how can I define the ROI?