Hi, it seems like after setting an interval which run an inference on every nth frame, the box is only drawn on every 'n’th frame. However, I want the box to be maintained on that position until the next inference is run (when the new bbox information is given). Would there be any way to do it on source config file?
Thank you so much in advance.
We do not support drawing box on uninferred frame, why you want the box maintain same position, object may change the position.
Well, if you’re performing inference on 1 frame per second. It would near impossible to view the bounding boxes if its a 30fps video. So retaining the boxes would help visually.
in one second, the object postion may change, the box will not at the position of object.
Yes, I understand that. But let’s say, I have an application where the boxes have low velocity and do not significantly move much in an interval of 1 second. In such an application, it will be useful to draw the bounding boxes from previous frames on undetected frames.
This also draws another question, when we use drop-interval. Are the uninferenced frames still passed through the GStreamer pipeline without ObjMeta? If that’s the case, we can manually store the bounding boxes and populate the ObjMeta of uninferenced frames.
Yes, the uninferenced frames still passed through the GStreamer pipeline without ObjMeta.