DeNoising-MOT: Towards Multiple Object Tracking with Severe Occlusions

TL;DR

DeNoising-MOT introduces a robust end-to-end transformer-based approach for multiple object tracking that effectively handles severe occlusions by simulating occlusion scenarios and employing a denoising training strategy, outperforming previous methods.

Contribution

The paper proposes a novel DeNoising Transformer architecture with a cascaded mask strategy, explicitly trained to handle occlusions without extra inference modules, advancing MOT performance in crowded scenes.

Findings

Outperforms state-of-the-art on MOT17, MOT20, and DanceTrack datasets.

Demonstrates robustness to severe occlusions and crowded scenes.

No additional modules needed during inference.

Abstract

Multiple object tracking (MOT) tends to become more challenging when severe occlusions occur. In this paper, we analyze the limitations of traditional Convolutional Neural Network-based methods and Transformer-based methods in handling occlusions and propose DNMOT, an end-to-end trainable DeNoising Transformer for MOT. To address the challenge of occlusions, we explicitly simulate the scenarios when occlusions occur. Specifically, we augment the trajectory with noises during training and make our model learn the denoising process in an encoder-decoder architecture, so that our model can exhibit strong robustness and perform well under crowded scenes. Additionally, we propose a Cascaded Mask strategy to better coordinate the interaction between different types of queries in the decoder to prevent the mutual suppression between neighboring trajectories under crowded scenes. Notably, the proposed method requires no additional modules like matching strategy and motion state estimation in inference. We conduct extensive experiments on the MOT17, MOT20, and DanceTrack datasets, and the experimental results show that our method outperforms previous state-of-the-art methods by a clear margin.