Secrets in Computing Optical Flow by Convolutional Networks

TL;DR

This paper explores the use of convolutional neural networks for optical flow estimation, proposing new architectures that reveal the intrinsic differences between various CNN structures for this task.

Contribution

It introduces novel CNN architectures specifically designed for optical flow estimation and analyzes their intrinsic differences.

Findings

Proposed CNN architectures outperform traditional methods.

Identified key structural differences affecting optical flow accuracy.

Demonstrated CNNs' potential in per-pixel regression tasks.

Abstract

Convolutional neural networks (CNNs) have been widely used over many areas in compute vision. Especially in classification. Recently, FlowNet and several works on opti- cal estimation using CNNs shows the potential ability of CNNs in doing per-pixel regression. We proposed several CNNs network architectures that can estimate optical flow, and fully unveiled the intrinsic different between these structures.