Skip to content

README.md

Latest commit

 

History

History
59 lines (36 loc) · 2.32 KB

README.md

File metadata and controls

59 lines (36 loc) · 2.32 KB

CSSL-Event-Object-Detection

Citation

Shenqi Wang and Guangzhi Tang, "Context-aware Sparse Spatiotemporal Learning for Event-based Vision," 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025.

@inproceedings{wang2025context,
  author={Wang, Shenqi and Tang, Guangzhi},
  booktitle={2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
  title={Context-aware Sparse Spatiotemporal Learning for Event-based Vision}, 
  year={2025},
  pages={1-7}}

Dependencies

Prophesee Metavision SDK

We are using Metavision 4.2.1.

Install Prophesee Metavision SDK using this LINK

Prophesee 1 Megapixel Event Dataset

Download DAT dataset from HERE.

Follow the tutorial HERE to pre-process the DAT file to HDF5 file, the pre-processing method used in this project is multi_channel_timesurface.

The pre-comupted dataset can be downloaded HERE. However, we noticed that the content of DAT and pre-computed are different. Our results are trained and tested on the preprocessed DAT dataset.

Enviornment setup

python -m pip install -r requirements.txt

Training and Testing

Training

Please modify the dataset_path in the train.py file to your own dataset path.

Use command python train.py to run the training script.

Validation

Please modify the saved_model_path in the validate.py file to saved model directory.

Please modify the dataset_path in the validate.py file to your own dataset path.

Use command python validate.py to run the training script.

Test

Please modify the saved_model_path in the test.py file to saved model directory.

Select the best performance mode drom the validation and modify the test_epoch in test.py.

Please modify the dataset_path in the test.py file to your own dataset path.

Use command python test.py to run the training script.