Weihao issues

Repositories
Issues
Comments

Results 3 issues of


                                            Weihao

About the coefficient 1000.0 in DKD.py

in line 21 - 26 in DKD.py > pred_teacher_part2 = F.softmax( > logits_teacher / temperature - 1000.0 * gt_mask, dim=1 > ) > log_pred_student_part2 = F.log_softmax( > logits_student / temperature...

关于optimizer的问题

我发现STDC中的optimizer与BiSeNet的optimizer相比多传入了boundary_loss_func,输出看了下，请问是因为在训练过程中对初始设置的融合权重0.6,0.3,0.1进行优化吗？

How to split the datasets (cityscapes) into labeled and unlabeled?

Hi, nice work and I am going to reproduce the experimental results. I want to know how you split the dataset into the labeld and unlabeled parts? Is it random...