Domain-general Crowd Counting in Unseen Scenarios

Official PyTorch implementation of 'Domain-general Crowd Counting in Unseen Scenarios'.(AAAI 2023 Oral) [arXiv]

Preparation

Python ≥ 3.7.

To install the required packages, please run:

pip install -r requirements.txt

For the data preparation, we follow the processing code in C-3-Framework. (Find processed UCF-QNRF here)

Train

• Set up the settings in main.py
• Run 'main.py'

Evaluation

• Download our trained model, SHA or SHB.

• Modify the path to the dataset and model for evaluation in 'test.py'.

• Run 'test.py'

Acknowledgement

Part of codes are borrowed from DomainBed and dg_mmld. Thanks for their great work!

Citation

If you find this work useful, please cite

@article{du2022domain,
  title={Domain-general Crowd Counting in Unseen Scenarios},
  author={Du, Zhipeng and Deng, Jiankang and Shi, Miaojing},
  journal={arXiv preprint arXiv:2212.02573},
  year={2022}
}