FlowNet: Learning Optical Flow with Convolutional Networks

TL;DR

This paper introduces CNN architectures for optical flow estimation, utilizing synthetic data for training, and demonstrates their effectiveness and generalization to real datasets at real-time speeds.

Contribution

It presents novel CNN architectures for optical flow, including a correlation layer, and a synthetic dataset for training that generalizes well to real-world data.

Findings

Networks trained on synthetic data perform well on real datasets.

Achieves 5-10 fps with competitive accuracy.

Correlation layer improves flow estimation.

Abstract

Convolutional neural networks (CNNs) have recently been very successful in a variety of computer vision tasks, especially on those linked to recognition. Optical flow estimation has not been among the tasks where CNNs were successful. In this paper we construct appropriate CNNs which are capable of solving the optical flow estimation problem as a supervised learning task. We propose and compare two architectures: a generic architecture and another one including a layer that correlates feature vectors at different image locations. Since existing ground truth data sets are not sufficiently large to train a CNN, we generate a synthetic Flying Chairs dataset. We show that networks trained on this unrealistic data still generalize very well to existing datasets such as Sintel and KITTI, achieving competitive accuracy at frame rates of 5 to 10 fps.