Zhe Chen comments

Results 316 comments of


                                            Zhe Chen

About the GFlops on the ADE20K dataset.

oh, you should import mmcv_custom and mmseg_custom at the begining of this code

Poor performance of Mask2Former based on coco-stuff-164k

Hello, thank you for your feedback. Can you give me some information about your environment?

Poor performance of Mask2Former based on coco-stuff-164k

Could you please confirm whether the pre-trained weights have been successfully loaded?

Poor performance of Mask2Former based on coco-stuff-164k

I just checked the config, and I think one possible reason is that I ran this experiment using 2 nodes, resulting in a total batch size of 16. If you...

Error during detection Train!! TypeError: '<' not supported between instances of 'str' and 'int'

Thanks for your feedback. I have not encountered this problem. I guess it may be due to the environment. Can you give me more information about your environment?

about cropsize

Crop size 896 was first adopted in the SwinV2 paper, and in order to obtain higher mIoU performance, we also adopted this setting in some models to improve performance. SwinV2:...

In deformable attention, why sampling_offsets.bias is initialized as arithmetic progression and set to be no gradient

`sampling_offsets.bias` is not frozen during training, because the `no_grad` here will not take effect. About the initialization, in simple terms, this initialization is to place the sampling points on the...

When I try to replicate the code, the results are not satisfactory.

Hi tandangzuoren, I need some time to clean up the code of the ablation experiments.

import error, I already cd ops & sh make.sh # compile deformable attention

If you meet `dist_train.sh: Bad substitution` when using cmd `sh dist_train.sh`, you can try `bash dist_train.sh`.

Training VIT-Adapter on custom dataset

Hi, please train models using config with the `ss` suffix. For example, you should use `mask2former_beit_adapter_large_896_80k_ss` instead of `mask2former_beit_adapter_large_896_80k_ms`. `ms` means performing multi-scale testing, so it will be very slow.