Usage Guidelines

Please follow the steps below.

Preparation

We implement Co-DETR using MMDetection V3.3.0 and MMCV V2.1.0. The source code of MMdetection has been included in this repo. We test our models under python=3.8.20,pytorch=2.1.0,cuda=12.1. Other versions may not be compatible.

In a python virtual environment and in the root directory of the project , install packages in the requirements.txt file.

pip install -r requirements.txt

Training

Train Co-DINO with 4 GPUs:

cd mmdetection && sh tools/dist_train.sh my_configs/codino/codino_xxx.py 4

Test

The test folder contains two subfolders: test_data and test_result. Among them, test_data stores the test data, and test_result stores the test results.

For bar charts:

python chart_data_extract_bar.py

For scatter charts:

python chart_data_extract_scatter.py

You need to modify the path(s) in the file.