I test some Shelf images, the detecion result is bad.
Is the model you prove really well trained?
Or some thing wrong with the tool of model upgrading?
I just use tools/upgrade_model_version.py to upgrade the model weights file, and then edit configs/SKU_fusion_bfp_x101_32x4d.py to the style of V2 by hand.
my mmdet version is 2.7.0 ,
mmcv version is 1.1.5
I test some Shelf images, the detecion result is bad. Is the model you prove really well trained? Or some thing wrong with the tool of model upgrading?
I just use tools/upgrade_model_version.py to upgrade the model weights file, and then edit configs/SKU_fusion_bfp_x101_32x4d.py to the style of V2 by hand. my mmdet version is 2.7.0 , mmcv version is 1.1.5