search

tusen-ai / simpledet

A Simple and Versatile Framework for Object Detection and Instance Recognition
Apache License 2.0
3.08k stars 486 forks source link

Problem: Train on my own dataset. #227

Closed surserrr closed 5 years ago

surserrr commented 5 years ago

Thanks for your excellent work！ I want to train my own dataset using tridentnet. I convert my dataset to coco and run create_coco_roidb.py , change gpus and num_class in config. Do I need to change another parameters?

And I meet the problem when run
python detection_train.py --config config/tridentnet_r101v1c4_c5_1x.py

Traceback (most recent call last): File "detection_train.py", line 290, in train_net(parse_args()) File "detection_train.py", line 21, in train_net transform, data_name, label_name, metric_list = config.get_config(is_train=True) File "/media/fangjt/3dc196fc-1311-49bb-ad24-5f77d8735211/simpledet/config/tridentnet_r101v1c4_c5_1x.py", line 132, in get_config num_branch=Trident.num_branch, scaleaware=Trident.train_scaleaware) File "/media/fangjt/3dc196fc-1311-49bb-ad24-5f77d8735211/simpledet/models/tridentnet/builder.py", line 84, in get_train_symbol rpn_loss = rpn_head.get_loss(rpn_feat, rpn_cls_label, rpn_reg_target, rpn_reg_weight) File "/media/fangjt/3dc196fc-1311-49bb-ad24-5f77d8735211/simpledet/models/tridentnet/builder.py", line 310, in get_loss image_anchor = p.anchor_assign.image_anchor AttributeError: 'NoneType' object has no attribute 'image_anchor'

RogerChern commented 5 years ago

Sorry, this is a regression introduced by #224. Please change anchor_assign to anchor_generate

RogerChern commented 5 years ago

Close by #228