Learning Sparse auto-Encoders for Green AI image coding
Cyprien Gille, Frédéric Guyard, Marc Antonini, Michel Barlaud
Abstract
Recently, convolutional auto-encoders (CAE) were introduced for image coding. They achieved performance improvements over the state-of-the-art JPEG2000 method. However, these performances were obtained using massive CAEs featuring a large number of parameters and whose training required heavy computational power.In this paper, we address the problem of lossy image compression using a CAE with a small memory footprint and low computational power usage.In this work, we propose a constrained approach and a new structured sparse learning method. We design an algorithm and test it on three constraints: the classical ℓ<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> constraint, the ℓ<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1,∞</inf> and the new ℓ<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1,1</inf> constraint. Experimental results show that the ℓ<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1,1</inf> constraint provides the best structured sparsity, resulting in a high reduction of memory ( 82 %) and computational cost reduction (25 %), with similar rate-distortion performance as with dense networks.
BibTeX
@inproceedings{icassp2023_learningsparseau,
title = {Learning Sparse auto-Encoders for Green AI image coding},
author = {Cyprien Gille and Frédéric Guyard and Marc Antonini and Michel Barlaud},
booktitle = {ICASSP 2023},
year = {2023}
}