AttentivePrototypeSFUDA
AttentivePrototypeSFUDA copied to clipboard
deeptibhegde
Some question about self_training_utils.save_pseudo_label_epoch(model, target_loader, rank, prototype, leave_pbar=True, ps_label_dir=ps_label_dir, cur_epoch=cur_epoch)
Excuse me： I'm interested in your work of "Attentive Prototypes for Source-free Unsupervised Domain Adaptive 3D Object Detection", but I encountered an issue while reading the code. I would like to ask how to view files such as “AttentivePrototypeSFUDA/SECOND-iou/pcdet/utils/pycache/self_training_utils.cpython-37.pyc”. Specifically, could you also provide the detail code of pcdet/utils/self_training_utils.py in the project and send it to me? I emailed you, but did not receive your reply. Perhaps you are too busy. Looking forward to your reply. Thank you very much.
The pycache files are automatically generated and can be ignored. The self_training_utils.py will be updated in the codebase shortly.
Apologies, it looks like there was a problem with attaching the zipped file. I will resend it, please let me know if there are other issues.
Apologies, it looks like there was a problem with attaching the zipped file. I will resend it, please let me know if there are other issues.
Got it , Thanks very much!
Excuse me, I have some few questions to consult with you
Firstly, is the prototype_.npy file updated in both the source and target domains, or is it updated only in the target domain?
as in the file of rain_st_utils.py
loss, tb_dict, disp_dict = model_func(model, source_batch)
st_loss, st_tb_dict, st_disp_dict = model_func(model, target_batch)
When I use the target domain dataset for training, all values in the obtained prototype_.npy file are nan.
Secondly, will the Domain Adaptation Module be accessed during the inference phase or just be accessed in the training phase?
Finally, your updated version of the file has undergone significant refactoring. Could you please provide me with the corresponding pcdet/utils/self_training_utils.py files corresponding to this GitHub repository.?
Thanks again.