# Learning Optimal Parameters for Multi-target Tracking with Contextual Interactions

arXiv: https://arxiv.org/abs/1610.01394 · HTML: https://mlsanity.com/paper/1610.01394/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Oct 2016.

## Built on by (most cited first, 8 of 8)

- [Learnable Graph Matching: Incorporating Graph Partitioning with Deep Feature Learning for Multiple Object Tracking](https://mlsanity.com/paper/2103.16178/) (2021)
- [3D Multi-Object Tracking Using Graph Neural Networks with Cross-Edge Modality Attention](https://mlsanity.com/paper/2203.10926/) (2022)
- [Track without Appearance: Learn Box and Tracklet Embedding with Local and Global Motion Patterns for Vehicle Tracking](https://mlsanity.com/paper/2108.06029/) (2021)
- [TrackMPNN: A Message Passing Graph Neural Architecture for Multi-Object Tracking](https://mlsanity.com/paper/2101.04206/) (2021)
- [Split and Connect: A Universal Tracklet Booster for Multi-Object Tracking](https://mlsanity.com/paper/2105.02426/) (2021)
- [Learnable Graph Matching: A Practical Paradigm for Data Association](https://mlsanity.com/paper/2303.15414/) (2023)
- [SKoPe3D: A Synthetic Dataset for Vehicle Keypoint Perception in 3D from Traffic Monitoring Cameras](https://mlsanity.com/paper/2309.01324/) (2023)
- [S3MOT: Monocular 3D Object Tracking with Selective State Space Model](https://mlsanity.com/paper/2504.18068/) (2025)

## Abstract (arXiv)

We describe an end-to-end framework for learning parameters of min-cost flow multi-target tracking problem with quadratic trajectory interactions including suppression of overlapping tracks and contextual cues about cooccurrence of different objects. Our approach utilizes structured prediction with a tracking-specific loss function to learn the complete set of model parameters. In this learning framework, we evaluate two different approaches to finding an optimal set of tracks under a quadratic model objective, one based on an LP relaxation and the other based on novel greedy variants of dynamic programming that handle pairwise interactions. We find the greedy algorithms achieve almost equivalent accuracy to the LP relaxation while being up to 10x faster than a commercial LP solver. We evaluate trained models on three challenging benchmarks. Surprisingly, we find that with proper parameter learning, our simple data association model without explicit appearance/motion reasoning is able to achieve comparable or better accuracy than many state-of-the-art methods that use far more complex motion features or appearance affinity metric learning.