irvingao/ddod

Official Implementation of DDOD (Disentangle your Dense Object Detector), ACM MM2021 oral

Disentangle Your Dense Object Detector

This repo contains the supported code and configuration files to reproduce object detection results of Disentangle Your Dense Object Detector. It is based on mmdetection.

Results and Models

Model	Backbone	Lr Schd	box mAP	AP50	AP75	APs	APm	APl
ATSS(IoU)	ResNet50	1x	39.4	56.6	42.6	23.9	42.5	49.6
DDOD	ResNet50	1x	41.6	59.9	45.2	23.9	44.9	54.4
DDOD-FCOS	ResNet50	1x	41.6	59.9	45.3	24.0	44.6	54.8

Usage

Installation

Please refer to get_started.md for installation and dataset preparation.

Or you can also fllow the simple steps:

git clone https://github.com/Irvingao/DDOD.git
cd DDOD
pip install -r requirements/build.txt
pip install -v -e .  

Inference

# multi-gpu testing
tools/dist_test.sh coco_cfg/ddod_r50_1x.py <DET_CHECKPOINT_FILE> 8 --eval bbox

Training

To train a detector with pre-trained models, run:

# single-gpu training
python tools/train.py coco_cfg/ddod_r50_1x.py

# multi-gpu training
tools/dist_train.sh coco_cfg/ddod_r50_1x.py 8

Citing DDOD

@misc{chen2021disentangle,
      title={Disentangle Your Dense Object Detector}, 
      author={Zehui Chen and Chenhongyi Yang and Qiaofei Li and Feng Zhao and Zhengjun Zha and Feng Wu},
      year={2021},
      eprint={2107.02963},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}