I have implemented the camera x with a use case of analyzer. The FPS speed I am geeting is around 30 fps. Each frame is passed ONNX framework for object detection but it is taking time to detect around 900ms for each frame. So in the pipeline there are huge number of frames that are pending for detection.
How can I implement it as efficient as possible so that user can se get the AI result as fast as possible?
I am thinking of using pipe and filter design pattern, will it solve the issue?
NOTE: I cannot skip frames from the camera pipeline.
Thanks