Deep-LK for Efficient Adaptive Object Tracking

TL;DR

Deep-LK introduces a novel, efficient regression-based object tracking method inspired by classical algorithms, outperforming prior deep trackers in accuracy and achieving real-time performance.

Contribution

The paper establishes a theoretical link between regression networks and the IC-LK algorithm and proposes Deep-LK, a new adaptive tracking framework inspired by IC-LK.

Findings

Deep-LK outperforms GOTURN in accuracy.

Deep-LK achieves real-time tracking at 100 FPS.

Comparable performance to state-of-the-art deep trackers.

Abstract

In this paper we present a new approach for efficient regression based object tracking which we refer to as Deep- LK. Our approach is closely related to the Generic Object Tracking Using Regression Networks (GOTURN) framework of Held et al. We make the following contributions. First, we demonstrate that there is a theoretical relationship between siamese regression networks like GOTURN and the classical Inverse-Compositional Lucas & Kanade (IC-LK) algorithm. Further, we demonstrate that unlike GOTURN IC-LK adapts its regressor to the appearance of the currently tracked frame. We argue that this missing property in GOTURN can be attributed to its poor performance on unseen objects and/or viewpoints. Second, we propose a novel framework for object tracking - which we refer to as Deep-LK - that is inspired by the IC-LK framework. Finally, we show impressive results demonstrating that Deep-LK substantially outperforms GOTURN. Additionally, we demonstrate comparable tracking performance to current state of the art deep-trackers whilst being an order of magnitude (i.e. 100 FPS) computationally efficient.