/DeConv-FTA

The fast transformation algorithm for transposed convolutional layers

Primary LanguagePython

A Fast Transformation Algorithm for Deconvolution

It provides detailed transformation parameters of the paper, which proposes a fast transformation algorithm (FTA) to accelerate deconvolution computation for deep nerual networks, eg. Generative Adversarial Network, Fully Convolutional Networks.

Wendong Mao, Peixiang Yang and Zhongfeng Wang, "FTA-GAN: A Computation-Efficient Accelerator for GANs With Fast Transformation Algorithm" in IEEE Transactions on Neural Networks and Learning Systems: Regular Paper, 2021.

[Paper] | Detailed Parameters of FTA for various deconvolutions(/transposed convolutions) will be available soon.

Please consider citing our paper if you find it useful for your work.

Contact

Wendong Mao

wdmao@smail.nju.edu.cn