This folder contains the implementation of DGMN2 for object detection.
Here, we take MMDetection as an example, applying DGMN2 to RetinaNet and Mask R-CNN.
| Method | Backbone | Lr schd | box AP | mask AP | Config | Download |
|---|---|---|---|---|---|---|
| RetinaNet | DGMN2-Tiny | 1x | 39.7 | - | config | model |
| RetinaNet | DGMN2-Small | 1x | 42.5 | - | config | model |
| RetinaNet | DGMN2-Medium | 1x | 43.7 | - | config | model |
| RetinaNet | DGMN2-Large | 1x | 44.7 | - | config | model |
| Mask R-CNN | DGMN2-Tiny | 1x | 40.1 | 37.2 | config | model |
| Mask R-CNN | DGMN2-Small | 1x | 43.4 | 39.7 | config | model |
| Mask R-CNN | DGMN2-Medium | 1x | 44.4 | 40.2 | config | model |
| Mask R-CNN | DGMN2-Large | 1x | 46.2 | 41.6 | config | model |
Clone the repository locally:
git clone https://github.com/fudan-zvg/DGMN2
a. Install MMDetection following the official instructions. Here we use MMDetection 2.12.0.
b. Install PyTorch Image Models. Here we use PyTorch Image Models 0.4.5.
pip install timm==0.4.5
c. Build the extension.
cd dcn
python setup.py build_ext --inplace
First, prepare COCO dataset according to the guidelines in MMDetection.
Then, download the weights pretrained on ImageNet, and put them in a folder pretrained/.
To train DGMN2-Tiny + RetinaNet on COCO train2017 on a single node with 8 GPUs for 12 epochs run:
dist_train.sh configs/retinanet_dgmn2_tiny_fpn_8x2_1x_coco.py 8
To evaluate DGMN2-Tiny + RetinaNet on COCO val2017 on a single node with 8 GPUs run:
dist_test.sh configs/retinanet_dgmn2_tiny_fpn_8x2_1x_coco.py /path/to/checkpoint_file 8 --eval bbox