A Generalized Loss Function for Crowd Counting and Localization

Data preparation

The dataset can be constructed followed by Bayesian Loss.

Pretrained model

The pretrained model can be downloaded from GoogleDrive.

Test

python test.py --data-dir PATH_TO_DATASET --save-dir PATH_TO_CHECKPOINT

Train

python train.py --data-dir PATH_TO_DATASET --save-dir PATH_TO_CHECKPOINT

Citation

If you use our code or models in your research, please cite with:

@InProceedings{Wan_2021_CVPR,
    author    = {Wan, Jia and Liu, Ziquan and Chan, Antoni B.},
    title     = {A Generalized Loss Function for Crowd Counting and Localization},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    year      = {2021},
    pages     = {1974-1983}
}

Acknowledgement

We use GeomLoss package to compute transport matrix. Thanks for the authors for providing this fantastic tool. The code is slightly modified to adapt to our framework.