DiffJPEG: A PyTorch implementation

This is a pytorch implementation of differentiable jpeg compression algorithm. This work is based on the discussion in this paper. The work relies heavily on the tensorflow implementation in this repository

Requirements

• PyTorch
• NumPy

Installation

python -m pip install git+git://github.com/cxy1997/DiffJPEG.git

Usage

DiffJPEG functions as a standard pytorch module/layer. To use, first import the layer and then initialize with the desired parameters:

• differentaible(bool): If true uses custom differentiable rounding function, if false uses standrard torch.round
• quality(float): Quality factor for jpeg compression scheme.

from DiffJPEG import DiffJPEG
jpeg = DiffJPEG(height=224, width=224, differentiable=True, quality=80)