Any insights on how to run an OpenVX pipeline and encode its output with minimal or no copy, and especially with the encoders not having to wait for a complete frame? I wish the NVIDIA renderer supported directing the output through the h.265 encoder instead of just to the current display!
Please go to https://forums.khronos.org/forumdisplay.php/110-OpenVX-Portable-Power-efficient-Vision-Processing
It is 3rdparty implementation and we don’t have much experience in it. The encoder should be SW encoder.