Crowd-DCNet

Crowd Counting application based on S-DCNet & SS-DCNet

How To Use

• Run this to have all your dependencies installed pip3 install -r requirements.txt --user.
• Download the S-DCNet pretrained weights from Google Drive or the SS-DCNet from Here (SHA weights are the only one tested).
• Extract the models folder into the Repo directory.
• Run python3 demo.py <pretrained_weights> -v <video_path> to use the script.
• Choose ROIs by hitting Space or Enter after every selecton and when finished hit ESC (but be warned that this has a great hit on the speed).

Note: SS-DCNet will work by default, if the user wish to use S-DCNet just add --SDCNet to the python3 running command.

News

• 26 Jan 2020 Add Camera Support & Threading for faster fetching ...... you can add --cap then specify the camera number (default=0) and the script will work online (press q to Quit)

Future Work

• Add ROI (Region of Interest) feaure.
• Optimize the code for faster inference.
• Upgrade to SS-DCNet.
• Make a Dockerfile of the project for easy deployment.

✨ Huge thanks for the real heroes here✨

If you find this work or code useful for your research, please cite:

@inproceedings{xhp2019SDCNet,
  title={From Open Set to Closed Set: Counting Objects by Spatial Divide-and-Conquer},
  author={Xiong, Haipeng and Lu, Hao and Liu, Chengxin and Liang, Liu and Cao, Zhiguo and Shen, Chunhua},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  year={2019},
  pages = {8362-8371}
}