When I built and ran "main/face_detector.cpp" published in UnitV2Framework, the image update in "http://unitv2.py" took more than a few minutes and was not practical. The "face_detector" in the factory-shipped M5stack UnitV2 updates the image almost in real time.
What we are currently checking is that the value of "faceobjects.size()" on the 74th line of "face_detector.cpp" is large and the processing is expected to be heavy.
I would like to know how to solve this problem.