SCAR

The code for "SCAR: Spatial-/Channel-wise Attention Regression Networks for Crowd Counting"

This is an official implementation of the paper "SCAR: Spatial-/Channel-wise Attention Regression Networks for Crowd Counting" (completed in September 2018, accepted by Neurocomputing in August 2019).

The original implementation of our paper is similar to this framework: GCC-SFCN. Recently, we re-implement SCAR using C^3 Framework and achieve a better performance than the original implementation. Thus, this repo provide the key code and hyperparameters in C^3 Framework.

Performance on Shanghai Tech Part B (MAE/MSE)

| Method | MAE | MSE | |-----------------------|---------|----------| | The orginal paper | 9.5 | 15.2 | | C^3 Framework | 9.0 | 13.8 |

Citing

If you use the code, please cite the following paper:

@article{gao2019scar,
  title={SCAR: Spatial-/channel-wise attention regression networks for crowd counting},
  author={Gao, Junyu and Wang, Qi and Yuan, Yuan},
  journal={Neurocomputing},
  volume={363},
  pages={1--8},
  year={2019},
  publisher={Elsevier}
}