MSSRM

An implementation of "Super-Resolution Information Enhancement For Crowd Counting" (Accepted by ICASSP 2023)

Environment

python >=3.7
pytorch >=1.7
opencv-python >=4.0
scipy >=1.4.0
h5py >=2.10
pillow >=7.0.0

Datasets

Download SR-Crowd dataset from Baidu-Disk, password:mvi3 ; or Google-Drive

Model

Download the pretrained model from from Baidu-Disk, password:ma5g ; or Google-Drive

Quickly test

• git clone https://github.com/Xiejiahao233/MSSRM.git
cd MSSRM
  

• Download Dataset and Model
  

• Generate images list

  Edit "make_npydata.py" to change the path to your original dataset folder.
  Run python make_npydata.py .
  

• Test
  python val.py --test_dataset Crowdsr --pre ./model/Crowdsr/model_best.pth --gpu_id 0
  

Training

Coming soon.

Reference

@article{xie2023super,
  title={Super-Resolution Information Enhancement For Crowd Counting},
  author={Xie, Jiahao and Xu, Wei and Liang, Dingkang and Ma, Zhanyu and Liang, Kongming and Liu, Weidong and Wang, Rui and Jin, Ling},
  journal={arXiv preprint arXiv:2303.06925},
  year={2023}
}