Fine-Grained Crowd Counting

Data preparation

The dataset can be downloaded from GoogleDrive.

Train, val, and test split are stored in Fine-Grained-dataset/Application/annotations/*.json

Point annotations are stored in Fine-Grained-Dataset/Application/annotations/annotations.json

Density map should be generated and named according to the data loader.

Train and Test

Training and testing examples can be found in scripts folder.

Citation

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

@article{wan2021fine,
    title={Fine-grained crowd counting},
    author={Wan, Jia and Kumar, Nikil Senthil and Chan, Antoni B},
    journal={IEEE transactions on image processing},
    volume={30},
    pages={2114--2126},
    year={2021},
}