Question about the FGFA part

Scalsol / mega.pytorch

Memory Enhanced Global-Local Aggregation for Video Object Detection, CVPR2020

Other

565 stars 115 forks source link

Question about the FGFA part #54

Closed hf1995 closed 3 years ago

hf1995 commented 4 years ago

hi! I test your FGFA_R_101.pth weight. I set _C.MODEL.VID.FGFA.ALL_FRAME_INTERVAL = 3 and _C.MODEL.VID.FGFA.KEY_FRAME_LOCATION = 1, other parameters remain unchanged. The results are AP50=0.60, which is lower than the single frame baseline. This result is abnormal. Are there any other parameters that need to be modified?