Total Variation Optimization Layers for Computer Vision

TL;DR

This paper introduces total variation minimization as a new layer in deep neural networks for computer vision, demonstrating its effectiveness across multiple tasks and developing a fast GPU-based optimization method.

Contribution

It proposes using total variation as a layer in deep nets for vision tasks and introduces a highly efficient GPU-based optimization algorithm.

Findings

Improved performance on five computer vision tasks.

Developed a GPU-based projected-Newton method that is 37 times faster.

Validated the usefulness of TV layers in deep learning models.

Abstract

Optimization within a layer of a deep-net has emerged as a new direction for deep-net layer design. However, there are two main challenges when applying these layers to computer vision tasks: (a) which optimization problem within a layer is useful?; (b) how to ensure that computation within a layer remains efficient? To study question (a), in this work, we propose total variation (TV) minimization as a layer for computer vision. Motivated by the success of total variation in image processing, we hypothesize that TV as a layer provides useful inductive bias for deep-nets too. We study this hypothesis on five computer vision tasks: image classification, weakly supervised object localization, edge-preserving smoothing, edge detection, and image denoising, improving over existing baselines. To achieve these results we had to address question (b): we developed a GPU-based projected-Newton method which is $37\times$ faster than existing solutions.