Miao Zheng

Results 80 comments of Miao Zheng

There might be syntax error in your config file, and I just suggest you try to use syntax checking in your IDE.

I suggest you might try to set `workers_per_gpu=0` in cfg file.

I suggest doing some preprocessing to resize crop or pad the images and make the height and width of them divisible by 16

Based on the error log, there might be some bc-breaks caused by mmcv updates, and I just suggest opening an issue in mmcv.

I think the performance reduction is caused by removing resize, as pre-trained models in mmseg follow the data augmentation setting with resize and we test them with resize also https://github.com/open-mmlab/mmsegmentation/blob/dd42fa8d0125632371a41a87c20485494c973535/configs/_base_/datasets/ade20k.py#L10...

Here is a tutorial for dataset preparation https://github.com/open-mmlab/mmsegmentation/blob/master/docs/en/dataset_prepare.md#ade20k

We close this pr as it has been fix in mim. ref: https://github.com/open-mmlab/mim/pull/160, https://github.com/open-mmlab/mim/pull/161.

we have support Migration Guides [here](https://github.com/open-mmlab/mmsegmentation/tree/1.x/docs/en/migration)

This issue is inactive for a long time, and I'll close it. If anyone has any questions about it, just reopen it.