Pytorch implementation of DifferentialConvolution.
Differential convolution is a plug-and-play module considering directional differences in spatial information to boost convolution efficiency.
Paper: https://www.sciencedirect.com/science/article/pii/S0893608019301315
Citation:
@article{sarigul2019differential,
title={Differential convolutional neural network},
author={Sar{\i}g{\"u}l, Mehmet and Ozyildirim, Buse Melis and Avci, Mutlu},
journal={Neural Networks},
volume={116},
pages={279--287},
year={2019},
publisher={Elsevier}
}