search

tianzhi0549 / FCOS

FCOS: Fully Convolutional One-Stage Object Detection (ICCV'19)
https://arxiv.org/abs/1904.01355
Other
3.27k stars 629 forks source link

subsample()->prepare_targets() : for proposals_per_image, targets_per_image in zip(proposals, targets): TypeError: 'NoneType' #369

Open ZhuangTingT opened 2 years ago

ZhuangTingT commented 2 years ago

hello, error: subsample()->prepare_targets() : for proposals_per_image, targets_per_image in zip(proposals, targets): TypeError: 'NoneType' I met the error in fcos_core/modeling/roi_heads/box_head/loss.py line 55, it seems the targets were not input, i find the dataloader only load sigle dataset(in fcos_core/data/build.py def build_dataset()) during training. Is it the reason? How can i fix it?

ps: In fcos_core/modeling/roi_heads/box_head/roi_box_feature_extractors.py around line 28, I changed head = resnet.ResNetHead to head = resnet.ResNet(config), is it right? Does the change caused the error above?

thanks, waiting for your reply!

SXYSUOSUO commented 1 year ago

Hello! Have you solved it yet? I also encountered the same problem, the "proposals" is None