Joint Coarse-And-Fine Reasoning for Deep Optical Flow

TL;DR

This paper introduces a joint coarse-and-fine reasoning approach for deep optical flow estimation, combining classification and regression to improve accuracy and efficiency in pixel-wise predictions.

Contribution

It presents a novel CNN architecture that jointly estimates coarse and fine optical flow, enhancing accuracy and reducing training time compared to existing methods.

Findings

Outperforms state-of-the-art CNN-based optical flow methods

Achieves higher accuracy on large optical flow datasets

Reduces training time through joint reasoning approach

Abstract

We propose a novel representation for dense pixel-wise estimation tasks using CNNs that boosts accuracy and reduces training time, by explicitly exploiting joint coarse-and-fine reasoning. The coarse reasoning is performed over a discrete classification space to obtain a general rough solution, while the fine details of the solution are obtained over a continuous regression space. In our approach both components are jointly estimated, which proved to be beneficial for improving estimation accuracy. Additionally, we propose a new network architecture, which combines coarse and fine components by treating the fine estimation as a refinement built on top of the coarse solution, and therefore adding details to the general prediction. We apply our approach to the challenging problem of optical flow estimation and empirically validate it against state-of-the-art CNN-based solutions trained from scratch and tested on large optical flow datasets.