Learning Multi-target Tracking with Quadratic Object Interactions

TL;DR

This paper introduces a quadratic interaction model for multi-target tracking that incorporates pairwise object interactions, improving accuracy and efficiency over existing methods.

Contribution

It proposes a novel quadratic extension to the min-cost flow model for multi-target tracking, with a new greedy inference algorithm that is faster and performs comparably to LP relaxation.

Findings

Greedy algorithm achieves similar accuracy to LP relaxation.

Model outperforms existing methods on KITTI benchmark.

Greedy approach is 2-7x faster than commercial solvers.

Abstract

We describe a model for multi-target tracking based on associating collections of candidate detections across frames of a video. In order to model pairwise interactions between different tracks, such as suppression of overlapping tracks and contextual cues about co-occurence of different objects, we augment a standard min-cost flow objective with quadratic terms between detection variables. We learn the parameters of this model using structured prediction and a loss function which approximates the multi-target tracking accuracy. We evaluate two different approaches to finding an optimal set of tracks under model objective based on an LP relaxation and a novel greedy extension to dynamic programming that handles pairwise interactions. We find the greedy algorithm achieves equivalent performance to the LP relaxation while being 2-7x faster than a commercial solver. The resulting model with learned parameters outperforms existing methods across several categories on the KITTI tracking benchmark.