FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks

TLDR

Optical flow estimation has been reframed as a learning problem by FlowNet, yet its accuracy, especially for small displacements and real‑world data, still lags behind traditional variational methods. This work extends end‑to‑end learning of optical flow to achieve competitive performance. The authors improve quality and speed through a data‑presentation schedule, a stacked architecture with image warping, and a dedicated subnetwork for small motions. FlowNet 2.0 reduces estimation error by over 50 %, matches state‑of‑the‑art accuracy while running at interactive rates, and offers faster variants up to 140 fps with comparable precision.

Abstract

The FlowNet demonstrated that optical flow estimation can be cast as a learning problem. However, the state of the art with regard to the quality of the flow has still been defined by traditional methods. Particularly on small displacements and real-world data, FlowNet cannot compete with variational methods. In this paper, we advance the concept of end-to-end learning of optical flow and make it work really well. The large improvements in quality and speed are caused by three major contributions: first, we focus on the training data and show that the schedule of presenting data during training is very important. Second, we develop a stacked architecture that includes warping of the second image with intermediate optical flow. Third, we elaborate on small displacements by introducing a subnetwork specializing on small motions. FlowNet 2.0 is only marginally slower than the original FlowNet but decreases the estimation error by more than 50%. It performs on par with state-of-the-art methods, while running at interactive frame rates. Moreover, we present faster variants that allow optical flow computation at up to 140fps with accuracy matching the original FlowNet.