Create README.md

README.md

@@ -0,0 +1,38 @@
+## imTED
+
+Code of [Integrally Migrating Pre-trained Transformer Encoder-decoders for  Visual Object Detection](https://arxiv.org/abs/2205.09613).
+
+The Code is based on [mmdetection](https://github.com/open-mmlab/mmdetection), please refer to [get_started.md](docs/en/get_started.md) and [MMDET_README.md](MMDET_README.md) to set up the environment and prepare the data.
+
+## Config Files and Performance
+
+We provide 9 configuration files in the configs directory.
+
+| Config File                                                                        | Backbone    | Epochs    | Box AP      | Mask AP   |
+| :--------------------------------------------------------------------------------: | :---------: | :-------: | :---------: | :-------: |
+| configs/imted/imted_faster_rcnn_vit_small_3x_coco.py                               | ViT-S       | 36        | 48.2        |           |  
+| configs/imted/imted_faster_rcnn_vit_base_3x_coco.py                                | ViT-B       | 36        | 52.9        |           | 
+| configs/imted/imted_faster_rcnn_vit_large_3x_coco.py                               | ViT-L       | 36        | 55.4        |           | 
+| configs/imted/imted_mask_rcnn_vit_small_3x_coco.py                                 | ViT-S       | 36        | 48.7        | 42.7      | 
+| configs/imted/imted_mask_rcnn_vit_base_3x_coco.py                                  | ViT-B       | 36        | 53.3        | 46.4      |
+| configs/imted/imted_mask_rcnn_vit_large_3x_coco.py                                 | ViT-L       | 36        | 55.5        | 48.1      |
+| configs/imted/few_shot/imted_faster_rcnn_vit_base_2x_base_training_coco.py         | ViT-B       | 24        | 50.6        |           |
+| configs/imted/few_shot/imted_faster_rcnn_vit_base_2x_finetuning_10shot_coco.py     | ViT-B       | 108       | 22.5        |           |
+| configs/imted/few_shot/imted_faster_rcnn_vit_base_2x_finetuning_30shot_coco.py     | ViT-B       | 108       | 30.2        |           |
+
+## MAE pre-training
+
+The pre-trained model is trained with the [official MAE code](https://github.com/facebookresearch/mae). 
+For ViT-S, we use a 4-layer decoder with dimension 256 for 800 epochs of pre-training. 
+For ViT-B, we use an 8-layer decoder with dimension 512 for 1600 epochs of pre-training. Pre-trained weights can be downloaded from the [official weight](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_base_full.pth).
+For ViT-L, we use an 8-layer decoder with dimension 512 for 1600 epochs of pre-training. Pre-trained weights can be downloaded from the[official weight](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_large_full.pth).
+
+For all experiments, remember to modify the path of pre-trained weights in the configuration files, e.g. configs/imted/imted_faster_rcnn_vit_small_3x_coco.py.
+
+For few-shot experiments, please refer to [FsDet](https://github.com/ucbdrive/few-shot-object-detection/blob/master/datasets/README.md#:~:text=2%2C%20and%203.-,COCO%3A,-cocosplit/%0A%20%20datasplit/%0A%20%20%20%20trainvalno5k) for data preparation. Remember to modify the path of json in the configuration files, e.g. configs/imted/few_shot/imted_faster_rcnn_vit_base_2x_base_training_coco.py.
+
+## Training with 8 GPUs
+
+```bash
+tools/dist_train.sh "path/to/config/file.py" 8
+```