谢昕辰

It depends on the network you used, most networks support different input sizes.

Hi @endaoguansanlu, It seems that mmcv-full is not properly installed, you could try to reinstall mmcv-full with the following command. ```shell pip uninstall mmcv pip install openmim mim install mmcv-full...

Could you show the codes you modified if possible? The scope is usually registered only once.

Hi @chen-del, Do you mean the config file or the model weights file? The config file generated during training contains the full config dict.

Hi, I think there is no need to modify the source code of the decoder if it only replaces the backbone. You might try to modify the [base config file](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/_base_/models/pspnet_r50-d8.py),...

Sorry... I'm not quite sure what your problem is. And we provide many backbone replacement examples, you might read them as reference. https://github.com/open-mmlab/mmsegmentation/blob/master/configs/_base_/models/fcn_unet_s5-d16.py https://github.com/open-mmlab/mmsegmentation/blob/master/configs/_base_/models/fcn_hr18.py https://github.com/open-mmlab/mmsegmentation/blob/master/configs/_base_/models/fcn_r50-d8.py

Hi @SBCV, Thanks for your bug report, you might modify the following code to solve this problem simply. And if you have a better solution, welcome to create a PR....

Hi @monkeycc, Could you describe your problem or requirements in detail(Datasets, usage scenarios, etc.)? According to the current description, we do not implement such a model, data augmentation method, or...

I think there are many ways to implement this method, but because of the lack of developers, we do not yet have plans to support this post-processing. You might try...

Hi @nicekwow, We have not yet implemented this feature due to a lack of developers, and we appreciate it if you submit a PR to add this feature. :)