Skip to content

Commit

Permalink
update README.md
Browse files Browse the repository at this point in the history
  • Loading branch information
@czczup
czczup committed May 31, 2022
1 parent 833121d commit 0716403
Showing 1 changed file with 10 additions and 10 deletions.
20 changes: 10 additions & 10 deletions detection/configs/upgraded_mask_rcnn/README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -8,20 +8,20 @@

Relative to the original Mask R-CNN, Li et al. modernized several of its modules. Concisely, the modifications include:

1. following the convolutions in FPN with batch normalization (BN)
1. following the convolutions in FPN with batch normalization (BN).

2. using two convolutional layers in the region proposal network (RPN) instead of one,
2. using two convolutional layers in the region proposal network (RPN) instead of one.

3. using four convolutional layers with BN followed by one linear layer for the region-of-interest (RoI) classification and box regression head instead of a two-layer MLP without normalization
3. using four convolutional layers with BN followed by one linear layer for the region-of-interest (RoI) classification and box regression head instead of a two-layer MLP without normalization.

4. following the convolutions in the standard mask head with BN.

## Results and Models

| Backbone | Pre-train | Lr schd | box AP | mask AP | #Param | Config | Download |
|:-------------:|:----------------------------------------------------------------------------------------------------------------------------------------------------------------:|:-------:|:------:|:-------:|:------:|:--------------------------------------------------------------:|:-------------------------------------------------------------------------------------------------------------------------------:|
| ViT-Adapter-T | [DeiT-T](https://dl.fbaipublicfiles.com/deit/deit_tiny_patch16_224-a1311bcf.pth) | 3x | 46.0 | 41.0 | 28M | [config](./mask_rcnn_deit_adapter_tiny_fpn_3x_coco.py) | [model](https://github.com/czczup/ViT-Adapter/releases/download/v0.1.2/mask_rcnn_deit_adapter_tiny_fpn_3x_coco.pth.tar) |
| ViT-Adapter-S | [DeiT-S](https://dl.fbaipublicfiles.com/deit/deit_small_patch16_224-cd65a155.pth) | 3x | 48.2 | 42.8 | 48M | [config](./mask_rcnn_deit_adapter_small_fpn_3x_coco.py) | [model](https://github.com/czczup/ViT-Adapter/releases/download/v0.1.2/mask_rcnn_deit_adapter_small_fpn_3x_coco.pth.tar) |
| ViT-Adapter-B | [DeiT-B](https://dl.fbaipublicfiles.com/deit/deit_base_patch16_224-b5f2ef4d.pth) | 3x | 49.6 | 43.6 | 120M | [config](./mask_rcnn_deit_adapter_base_fpn_3x_coco.py) | [model]() |
| ViT-Adapter-B | [Uni-Perceiver](https://github.com/czczup/ViT-Adapter/releases/download/v0.1.1/uniperceiver_pretrain.pth) | 3x | 50.7 | 44.9 | 120M | [config](./mask_rcnn_uniperceiver_adapter_base_fpn_3x_coco.py) | [model](https://github.com/czczup/ViT-Adapter/releases/download/v0.1.1/mask_rcnn_uniperceiver_adapter_base_fpn_3x_coco.pth.tar) |
| ViT-Adapter-L | [AugReg](https://storage.googleapis.com/vit_models/augreg/L_16-i21k-300ep-lr_0.001-aug_medium1-wd_0.1-do_0.1-sd_0.1--imagenet2012-steps_20k-lr_0.01-res_384.npz) | 3x | 50.9 | 44.8 | 348M | [config](./mask_rcnn_augreg_adapter_large_fpn_3x_coco.py) | [model](https://github.com/czczup/ViT-Adapter/releases/download/v0.1.2/mask_rcnn_augreg_adapter_large_fpn_3x_coco.pth.tar) |
| Backbone | Pre-train | Lr schd | box AP | mask AP | #Param | Config | Download |
|:-------------:|:----------------------------------------------------------------------------:|:-------:|:------:|:-------:|:------:|:-----------------------------------------------------------:|:----------------------------------------------------------------------------------------------------------------------------:|
| ViT-B | [MAE](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_base.pth) | 25ep | 48.1 | 43.2 | 116M | - | - |
| ViT-B | [MAE](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_base.pth) | 50ep | 50.1 | 44.6 | 116M | - | - |
| ViT-B | [MAE](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_base.pth) | 100ep | 50.3 | 44.9 | 116M | - | - |
| ViT-Adapter-B | [MAE](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_base.pth) | 25ep | 50.3 | 44.7 | 122M | [config](./mask_rcnn_mae_adapter_base_lsj_fpn_25ep_coco.py) | [model](https://github.com/czczup/ViT-Adapter/releases/download/v0.1.4/mask_rcnn_mae_adapter_base_lsj_fpn_25ep_coco.pth.tar) |
| ViT-Adapter-B | [MAE](https://dl.fbaipublicfiles.com/mae/pretrain/mae_pretrain_vit_base.pth) | 50ep | 50.8 | 45.1 | 122M | [config](./mask_rcnn_mae_adapter_base_lsj_fpn_50ep_coco.py) | [model](https://github.com/czczup/ViT-Adapter/releases/download/v0.1.4/mask_rcnn_mae_adapter_base_lsj_fpn_50ep_coco.pth.tar) |

0 comments on commit 0716403

Please sign in to comment.