ETTrack: Enhanced Temporal Motion Predictor for Multi-Object Tracking

TL;DR

This work proposes ETTrack, a novel motion prediction method with an enhanced temporal motion predictor that provides a robust solution for MOT in complex dynamic environments, which enhances the non-linear motion prediction capabilities of tracking algorithms.

Abstract

Many Multi-Object Tracking (MOT) approaches exploit motion information to associate all the detected objects across frames. However, many methods that rely on filtering-based algorithms, such as the Kalman Filter, often work well in linear motion scenarios but struggle to accurately predict the locations of objects undergoing complex and non-linear movements. To tackle these scenarios, we propose a motion-based MOT approach with an enhanced temporal motion predictor, ETTrack. Specifically, the motion predictor integrates a transformer model and a Temporal Convolutional Network (TCN) to capture short-term and long-term motion patterns, and it predicts the future motion of individual objects based on the historical motion information. Additionally, we propose a novel Momentum Correction Loss function that provides additional information regarding the motion direction of objects during training. This allows the motion predictor rapidly adapt to motion variations and more accurately predict future motion. Our experimental results demonstrate that ETTrack achieves a competitive performance compared with state-of-the-art trackers on DanceTrack and SportsMOT, scoring 56.4% and 74.4% in HOTA metrics, respectively.

Figures (7)

• Figure 1: The pipeline of ETTrack. The historical trajectory of length $p$ is fed into our motion predictor, which predicts the track predictions $P_s$ at the current moment $s$. By deploying an object detector, detections $D_s$ are obtained. With the track predictions and detections, data association can be accomplished by IoU Distance and Hungarian matching algorithm. Different colors represent different identities in the tracking results
• Figure 2: The network structure of Temporal Transformer in the motion predictor
• Figure 3: Structure of the TCN
• Figure 4: visualization of a stack of dilated causal convolutions ($K$ = 7, Dilation = [$\mathbf{2}^\mathbf{0}$,$\mathbf{2}^\mathbf{1}$,$\mathbf{2}^\mathbf{2}$,$\mathbf{2}^\mathbf{3}$])
• Figure 5: Illustration of momentum correction loss. We obtain the predicted and actual motion directions, which are represented by the blue and green arrows, respectively. Subsequently, we calculate the angular difference between the predicted and actual motion directions to obtain the momentum correction loss