TCSVT-LightWeight-CNNs

Code for our IEEE TCSVT Paper: Lightweight Modules for Efficient Deep Learning based Image Restoration

Authors: Avisek Lahiri*, Sourav Bairagya*, Sutanu Bera, Siddhant Haldar, Prabir Kumar Biswas
(* equal contribution)

1. Paper Link: https://arxiv.org/abs/2007.05835
   
2. IEEE Early Access Link: https://ieeexplore.ieee.org/document/9134805
   

Key Points from Paper

• Paper provides re-usable modules to be plugged and played to compress a given CNN
• Select any favourite full-scale baseline for low-level vision applications
• Replace 3X3 conv by LIST layer
• Replace dilated conv layer GSAT layer
• Achieve efficient up/down-sample with Bilinear SubSampling followed by LIST layer

TensorflowExamples

This contains the basic proposed modules in Tensorflow
TensorflowExamples/basicModules.py contains the proposed LIST, GSAT modules
It also contains the framework for LIST based up/down-sampling in a CNN

PytorchExamples

It contains the basic proposed modules in Pytorch

• block.py has the implementatin of LIST module based DnCNN denoising framework
• Denoising_demo.ipynb is a notebook to reflect our training/inference setup for DnCNN experiments