Simple Transformer with Single Leaky Neuron for Event Vision

Code for the paper: Simple Transformer with Single Leaky Neuron for Event Vision, accepted at EVGEN: Event-based Vision in the Era of Generative AI - Transforming Perception and Visual Innovation Workshop, WACV 2025.

---

Table of Contents

• Installation
• Dataset Setup
• Executing Program
• Citation
• Acknowledgements

Installation

1. Install spikingjelly from Github only.
2. Install rest of the packages as mentioned in requirements.txt .

Dataset Setup

1. Download the datasets manually from the spikingjelly datasets repository:
   • DVS Gesture
   • N-MNIST
   • CIFAR10-DVS
2. Place the downloaded datasets in the event_vision/datasets/ folder. The code will automatically preprocess the datasets into frame-based versions during execution.

Executing Program

• Download pretrained Resnet18 and Resnet50 weights from pytorch.
• Update the dataset paths in the custom_train.py file to match your setup.
• To start training, run the following command:

  python custom_train.py
• Logs and training progress will be stored in the logs/ folder.

Citation

If you find this code useful in your research, please consider citing our paper:

@InProceedings{Kumar_2025_WACV,
    author    = {Kumar, Himanshu and Konkar, Aniket},
    title     = {Simple Transformer with Single Leaky Neuron for Event Vision},
    booktitle = {Proceedings of the Winter Conference on Applications of Computer Vision (WACV) Workshops},
    month     = {February},
    year      = {2025},
    pages     = {928-934}
}

---

Acknowledgements

We have used Parametric-Leaky-Integrate-and-Fire-Spiking-Neuron code, and syops-counter for energy calculation.