MambaST

Xiangbo Gao, Asiegbu Miracle Kanu-Asiegbu, Xiaoxiao Du

MambaST: A Plug-and-Play Cross-Spectral Spatial-Temporal Fuser for Efficient Pedestrian Detection

Usage

1. Build Docker Image

   cd docker
   bash build.sh

   Note: Make sure to change USERNAME and USER_UID in the script.

2. Run Docker Container

   cd ..
   bash docker/run.sh

   Note: Update your_username and path_to_project_directory accordingly.

3. Install Third Party Libraries

   cd thirdparty/Vim
   pip install -e mamba-1p1p1
   cd ../..
   pip install causal-conv1d==1.1.0

4. Download and Prepare Dataset

   • Download the KAIST-CVPR15 dataset from here and place it in your dataset directory.
   • Copy the dataset:

     cp -r ~/dataset/* path_to_your_dataset_directory

   • Copy the sanitized annotations format:

     cp sanitized_annotations_format_all path_to_your_dataset_directory

5. Update Annotation Path

   • Modify KAIST_ANNOTATION_PATH in utils/datasets_vid.py to the absolute path of sanitized_annotations_format_all.

6. Training

   bash train.sh

   Sample training code.

7. Testing

   bash test.sh

Citation

If you found repo useful, feel free to cite.

@article{gao2024mambast,
  title={MambaST: A Plug-and-Play Cross-Spectral Spatial-Temporal Fuser for Efficient Pedestrian Detection},
  author={Gao, Xiangbo and Kanu-Asiegbu, Asiegbu Miracle and Du, Xiaoxiao},
  journal={arXiv preprint arXiv:2408.01037},
  year={2024}
}