Sensor-Guided Optical Flow

TL;DR

This paper introduces a framework that enhances optical flow accuracy by integrating external sparse hints derived from depth sensors and algorithms, improving performance on known and unseen domains.

Contribution

It presents a novel method to guide optical flow networks using external cues, combining sensor data and algorithms to generate effective hints for better predictions.

Findings

Improved optical flow accuracy on benchmarks

Effective guidance in simulated and real conditions

Compatible with state-of-the-art flow networks

Abstract

This paper proposes a framework to guide an optical flow network with external cues to achieve superior accuracy either on known or unseen domains. Given the availability of sparse yet accurate optical flow hints from an external source, these are injected to modulate the correlation scores computed by a state-of-the-art optical flow network and guide it towards more accurate predictions. Although no real sensor can provide sparse flow hints, we show how these can be obtained by combining depth measurements from active sensors with geometry and hand-crafted optical flow algorithms, leading to accurate enough hints for our purpose. Experimental results with a state-of-the-art flow network on standard benchmarks support the effectiveness of our framework, both in simulated and real conditions.