[Feature]: Convert scripts and instructions on seg and det (#621)

* [Feature]: Add converting script

* [Feature]: Add instruction for det and seg

* [Feature]: Fix citation

* [Fix]: Use absolute path

* [Fix]: Change tree to blob

configs/selfsup/mae/README.md

@@ -142,6 +142,19 @@ methods that use only ImageNet-1K data. Transfer performance in downstream tasks
 </tbody>
 </table>
 
+## Evaluating MAE on Detection and Segmentation
+
+If you want to evaluate your model on detection or segmentation task, we provide a [script](https://github.com/open-mmlab/mmselfsup/blob/dev-1.x/tools/model_converters/mmcls2timm.py) to convert the model keys from MMClassification style to timm style.
+
+```sh
+cd $MMSELFSUP
+python tools/model_converters/mmcls2timm.py $src_ckpt $dst_ckpt
+```
+
+Then, using this converted ckpt, you can evaluate your model on detection task, following [Detectron2](https://github.com/facebookresearch/detectron2/tree/main/projects/ViTDet)，
+and on semantic segmentation task, following this [project](https://github.com/implus/mae_segmentation). Besides, using the unconverted ckpt, you can use
+[MMSegmentation](https://github.com/open-mmlab/mmsegmentation/tree/master/configs/mae) to evaluate your model.
+
 ## Citation
 
 ```bibtex
@@ -149,7 +162,7 @@ methods that use only ImageNet-1K data. Transfer performance in downstream tasks
   title={Masked Autoencoders Are Scalable Vision Learners},
   author={Kaiming He and Xinlei Chen and Saining Xie and Yanghao Li and
   Piotr Doll'ar and Ross B. Girshick},
-  journal={ArXiv},
+  journal={arXiv},
   year={2021}
 }
 ```

tools/model_converters/mmcls2timm.py

@@ -0,0 +1,86 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import argparse
+import os.path as osp
+from collections import OrderedDict
+from typing import Union
+
+import mmengine
+import torch
+from mmengine.runner.checkpoint import _load_checkpoint
+
+
+def convert_mmcls_to_timm(state_dict: Union[OrderedDict, dict]) -> OrderedDict:
+    """Convert keys in MMClassification pretrained vit models to timm tyle.
+
+    Args:
+        state_dict (Union[OrderedDict, dict]): The state dict of
+            MMClassification pretrained vit models.
+
+    Returns:
+        OrderedDict: The converted state dict.
+    """
+    # only keep the backbone weights and remove the backbone. prefix
+    state_dict = {
+        key.replace('backbone.', ''): value
+        for key, value in state_dict.items() if key.startswith('backbone.')
+    }
+
+    # replace projection with proj
+    state_dict = {
+        key.replace('projection', 'proj'): value
+        for key, value in state_dict.items()
+    }
+
+    # replace ffn.layers.0.0 with mlp.fc1
+    state_dict = {
+        key.replace('ffn.layers.0.0', 'mlp.fc1'): value
+        for key, value in state_dict.items()
+    }
+
+    # replace ffn.layers.1 with mlp.fc2
+    state_dict = {
+        key.replace('ffn.layers.1', 'mlp.fc2'): value
+        for key, value in state_dict.items()
+    }
+
+    # replace layers with blocks
+    state_dict = {
+        key.replace('layers', 'blocks'): value
+        for key, value in state_dict.items()
+    }
+
+    # replace ln with norm
+    state_dict = {
+        key.replace('ln', 'norm'): value
+        for key, value in state_dict.items()
+    }
+
+    # replace the last norm1 with norm
+    state_dict['norm.weight'] = state_dict.pop('norm1.weight')
+    state_dict['norm.bias'] = state_dict.pop('norm1.bias')
+
+    state_dict = OrderedDict({'model': state_dict})
+    return state_dict
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description='Convert keys in MMClassification '
+        'pretrained vit models to timm tyle')
+    parser.add_argument('src', help='src model path or url')
+    parser.add_argument('dst', help='save path')
+    args = parser.parse_args()
+
+    checkpoint = _load_checkpoint(args.src)
+    if 'state_dict' in checkpoint:
+        state_dict = checkpoint['state_dict']
+    else:
+        state_dict = checkpoint
+
+    state_dict = convert_mmcls_to_timm(state_dict)
+    mmengine.mkdir_or_exist(osp.dirname(args.dst))
+    torch.save(state_dict, args.dst)
+
+
+if __name__ == '__main__':
+    main()