AEDNet: Asynchronous Event Denoising with Spatial-Temporal Correlation among Irregular Data

Source code for AEDNet paper from ACMMM 2022

A novel asynchronous event denoising neural network directly consumes the correlation of the irregular signal in spatiotemporal range without destroying its original structural property.

If you find this work useful in your academic reaserch, please cite the following work:

@inproceedings{fang2022aednet,
  title={AEDNet: Asynchronous Event Denoising with Spatial-Temporal Correlation among Irregular Data},
  author={Fang, Huachen and Wu, Jinjian and Li, Leida and Hou, Junhui and Dong, Weisheng and Shi, Guangming},
  booktitle={Proceedings of the 30th ACM International Conference on Multimedia},
  pages={1427--1435},
  year={2022}
}

If there is any suggestion or questions, feel free to fire an issue to let us know. :)

Installation

This code was tested on an Ubuntu 20.04.1 system (i9-10920X CPU, 128GB RAM, and GeForce RTX 3090Ti GPU) running Python 3.7, Pytorch 1.12 and CUDAToolkit 11.3.1.

conda create --name aednet python=3.7
conda activate aednet
pip install numpy=1.21.6
pip install matplotlib=3.5.2
pip install tensorboardX
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

Train the model

We have released the trained model parameters in Releases. If you want to train your own model, you should first put your data in "data/AEDNetDataset_BA" and list the name of data in "training_file_name.txt". After that, you can train your model via:

python train_net.py --trainset training_file_name.txt --testset test_file_name.txt --nepoch 2000 --batchSize 8

Test the model

Put the model parameters in Releases to "models/BA_noise_removal_model" and test it via:

python test_net.py --shapename MAH00444_50{i} --x_frame 1280 --y_frame 720

DVSCLEAN Dataset

To download the dataset use:DVSCLEAN