search

hnuzhy / SSDA-YOLO

Codes for my paper "SSDA-YOLO: Semi-supervised Domain Adaptive YOLO for Cross-Domain Object Detection"
134 stars 17 forks source link

Labes problem #19

Open DuyunliangToon opened 11 months ago

DuyunliangToon commented 11 months ago

Hi, I have a question about tags while reproducing your paper. According to the description in your dataset configuration file, train_source_real, train_source_fake and test_traget_real require labels, and train_target_real and train_target_fake do not require labels. But in my actual training, I found that train_source_fake does not need labels, and train_target_real needs labels. From this point of view, it requires all the labels of the two data sets, which is not the so-called semi-supervised training. Not sure if I made a mistake, hope you have time to help me out. image

hnuzhy commented 11 months ago

Yes, we do use labels of images in train_source_real and train_source_fake, and not use labels of images in train_target_real and train_target_fake. When applying dataloading, the train_source_fake has exactly the same labels as train_source_real, thus we do not need load its labels one more time. As for labels of train_target_real and train_target_fake, we do not use them when training, You can refer the training code in lines list below https://github.com/hnuzhy/SSDA-YOLO/blob/master/ssda_yolov5_train.py#L489~#L534 We want these labels in target domain just for training of Oracle experiments, which are fully supervised training and testing of the target domain.

DuyunliangToon commented 11 months ago

Thanks for your guidance.