InsightFace-REST
InsightFace-REST copied to clipboard
close faces not detected
trafficstars
Faces that take up the majority of the image are not detected. What parameters can I change to fix this?
Hi! What max_size and det_model arguments are currently used?
This issue mostly influence original scrfd_*g_bnkps models at higher resolution (640x640 and above), to partially solve this issue scrfd_*g_gnkps models were trained, which are now used by default.
Yolov5 family models mostly perform better in such scenario, though from time to time can miss some faces too.