Can't draw inference results using the OSD binary for deepstream_pose_estimation when convert format to RGBA before demuxer

Do you mean that there are changes other than adding processes to ignore feature points that are outside the frame of the video buffer?