Binhui Xie (谢斌辉)

Hi @extreme-assistant We request to add our new work RIPU to this awesome repo. Title: Towards Fewer Annotations: Active Learning via Region Impurity and Prediction Uncertainty for Domain Adaptive Semantic...

About sigmoid_focal_loss

5

Hello, guys. Well done! I have a quick question about `sigmoid_focal_loss` as follows: https://github.com/IDEACVR/DINO/blob/67bbcd97ef30a48cf343b7b0f3ad9ea0795b6fcd/models/dino/dino.py#L379 Why is the third dimension of `target_classes_onehot` one more than that of `src_logits`? Does the extra...

Inconsistent results in Table 3 (SYNTHIA->Cityscapes)

2

Hello, thank for sharing the great work. I find some inconsistent results in Table 3. Why the reported results of the CAG-UDA model in two mIoU metrics (13- and 16-class...

About pre-trained model

1

Can you provide VGG-16 based pre-trained model?

Hi @shaofeifei11, excellent work! Recently, we are replicating the results of S3DIS. However, we encountered some problems, that is, the training time is particularly long, the cpu occupies a lot...