TJEvents: A general deep learning based event camera data processing toolbox

This repo contains various general tools for event camera data processing as well as many general modules or models for deep learning on event camera, especially for graph-based methods or video reconstruction from events.

Now we have implemented:

The Event2Graph Dataset from the paper Graph-Based Object Classification for Neuromorphic Vision Sensing, based on the code.
The FireNet from the paper Fast Image Reconstruction with an Event Camera, based on the code.
The event data voxel grid representations from the paper High Speed and High Dynamic Range Video with an Event Camera, based on the code.

Requirements

torch >= 1.4.0
torchvision >= 0.5.0
torch_geometric >= 1.5.0
opencv-python
scipy

Install

a. Install PyTorch 1.5 and torchvision 0.6 following the official instructions.

b. Install Pytorch Geometric 1.5 following the official instructions.

c. Clone the repository.

git clone https://github.com/Flawless1202/tjevents.git

d. Install the other requirements.

pip install -r requirements.txt

d. Install.

python(3) setup.py install  # add --user if you want to install it locally
# or "pip install ."

Run a train example

a. Prepare the dataset: Download the ASL-Dataset and unzip them all to data/NVS2Graph/raw.

b. Run the train example with spcific config

python examples/train_e2g.py --load_config ./config/dgcnn.yaml