The dataset can be constructed followed by Bayesian Loss.
The pretrained model can be downloaded from GoogleDrive.
python test.py --data-dir PATH_TO_DATASET --save-dir PATH_TO_CHECKPOINT
python train.py --data-dir PATH_TO_DATASET --save-dir PATH_TO_CHECKPOINT
If you use our code or models in your research, please cite with:
@article{wan2020modeling,
title={Modeling Noisy Annotations for Crowd Counting},
author={Wan, Jia and Chan, Antoni},
journal={Advances in Neural Information Processing Systems},
volume={33},
year={2020}
}