Siamese Instance Search for Tracking

TL;DR

This paper introduces a novel tracking method called SINT that uses a Siamese neural network to match the initial target with candidates in new frames, achieving state-of-the-art results without model updates or occlusion handling.

Contribution

The paper presents a simple yet effective tracker based on a learned matching function that does not require target-specific data or online adaptation.

Findings

Achieves state-of-the-art tracking performance on OTB and challenging YouTube videos.

Does not require model updating, occlusion detection, or geometric matching.

Enables target re-identification after absence.

Abstract

In this paper we present a tracker, which is radically different from state-of-the-art trackers: we apply no model updating, no occlusion detection, no combination of trackers, no geometric matching, and still deliver state-of-the-art tracking performance, as demonstrated on the popular online tracking benchmark (OTB) and six very challenging YouTube videos. The presented tracker simply matches the initial patch of the target in the first frame with candidates in a new frame and returns the most similar patch by a learned matching function. The strength of the matching function comes from being extensively trained generically, i.e., without any data of the target, using a Siamese deep neural network, which we design for tracking. Once learned, the matching function is used as is, without any adapting, to track previously unseen targets. It turns out that the learned matching function is so powerful that a simple tracker built upon it, coined Siamese INstance search Tracker, SINT, which only uses the original observation of the target from the first frame, suffices to reach state-of-the-art performance. Further, we show the proposed tracker even allows for target re-identification after the target was absent for a complete video shot.