[Project] Added a supported for Visual Attention Network (VAN) #2987
Conversation
|
|
|
Hi @tang576225574, thank you for your nice pr, but please sign the CLA first before we can review on it. Besides, can you also fix the lint please, thank you so much. |
ok, I will fix it today. |
lint had been fixed. |
|
Hi @tang576225574, |
| pad_val=0, | ||
| seg_pad_val=255, | ||
| size=(512, 512), | ||
| test_cfg=dict(size_divisor=32)) |
There was a problem hiding this comment.
Might remove this setting, since the official implementation uses Resize instead of Padding.
https://github.com/Visual-Attention-Network/VAN-Segmentation/blob/main/configs/_base_/datasets/ade20k.py#L28
There was a problem hiding this comment.
Might remote this setting, since the official implementation uses
Resizeinstead ofPadding. https://github.com/Visual-Attention-Network/VAN-Segmentation/blob/main/configs/_base_/datasets/ade20k.py#L28
How to align this setting?
该怎么做？
There was a problem hiding this comment.
可以在你的project的config目录下新建一个ade20k.py，重新写一个test_pipeline。
There was a problem hiding this comment.
可以在你的project的config目录下新建一个ade20k.py，重新写一个test_pipeline。
ok....
of course! I will write a README in the next commit, about 3 days. |
|
Fixed！ |
xiexinch
changed the title
…mmlab#2987) ## Motivation The original version of Visual Attention Network (VAN) can be found from https://github.com/Visual-Attention-Network/VAN-Segmentation 添加Visual Attention Network (VAN)的支持。 ## Modification added a floder mmsegmentation/projects/van/ added 13 configs totally and aligned performance basically. 只增加了一个文件夹，共增加13个配置文件，基本对齐性能（没有全部跑）。 ## Use cases (Optional) Before running, you may need to download the pretrain model from https://cloud.tsinghua.edu.cn/d/0100f0cea37d41ba8d08/ and then move them to the folder mmsegmentation/pretrained/, i.e. "mmsegmentation/pretrained/van_b2.pth". After that, run the following command: cd mmsegmentation bash tools/dist_train.sh projects/van/configs/van/van-b2_pre1k_upernet_4xb2-160k_ade20k-512x512.py 4 --------- Co-authored-by: xiexinch <[email protected]>
Motivation
The original version of Visual Attention Network (VAN) can be found from
https://github.com/Visual-Attention-Network/VAN-Segmentation
添加Visual Attention Network (VAN)的支持。
Modification
added a floder mmsegmentation/projects/van/
added 13 configs totally and aligned performance basically.
只增加了一个文件夹，共增加13个配置文件，基本对齐性能（没有全部跑）。
Use cases (Optional)
Before running, you may need to download the pretrain model from https://cloud.tsinghua.edu.cn/d/0100f0cea37d41ba8d08/
and then move them to the folder mmsegmentation/pretrained/, i.e. "mmsegmentation/pretrained/van_b2.pth".
After that, run the following command:
cd mmsegmentation
bash tools/dist_train.sh projects/van/configs/van/van-b2_pre1k_upernet_4xb2-160k_ade20k-512x512.py 4