In Multiple Object Tracking, objects often exhibit non-linear motion of
acceleration and deceleration, with irregular direction changes.
Tacking-by-detection (TBD) with Kalman Filter motion prediction works well in
pedestrian-dominant scenarios but falls short in complex situations when
multiple objects perform non-linear and diverse motion simultaneously. To
tackle the complex non-linear motion, we propose a real-time diffusion-based
MOT approach named DiffMOT. Specifically, for the motion predictor component,
we propose a novel Decoupled Diffusion-based Motion Predictor (D MP). It models
the entire distribution of various motion presented by the data as a whole. It
also predicts an individual object's motion conditioning on an individual's
historical motion information. Furthermore, it optimizes the diffusion process
with much less sampling steps. As a MOT tracker, the DiffMOT is real-time at
22.7FPS, and also outperforms the state-of-the-art on DanceTrack and SportsMOT
datasets with 63.4 and 76.2 in HOTA metrics, respectively. To the best of our
knowledge, DiffMOT is the first to introduce a diffusion probabilistic model
into the MOT to tackle non-linear motion prediction.

非线性运动的多目标跟踪中引入扩散概率模型的 DiffMOT 方法能够更准确地预测目标的运动并取得较优的跟踪性能。

DiffMOT：一种基于扩散的实时多目标跟踪器与非线性预测

DiffMOT: A Real-time Diffusion-based Multiple Object Tracker with  Non-linear Prediction

Significant advancements have been made in multi-object tracking (MOT) with
the development of detection and re-identification (ReID) techniques. Despite
these developments, the task of accurately tracking objects in scenarios with
homogeneous appearance and heterogeneous motion remains challenging due to the
insufficient discriminability of ReID features and the predominant use of
linear motion models in MOT. In this context, we present a novel learnable
motion predictor, named MotionTrack, which comprehensively incorporates two
levels of granularity of motion features to enhance the modeling of temporal
dynamics and facilitate accurate future motion prediction of individual
objects. Specifically, the proposed approach adopts a self-attention mechanism
to capture token-level information and a Dynamic MLP layer to model
channel-level features. MotionTrack is a simple, online tracking approach. Our
experimental results demonstrate that MotionTrack yields state-of-the-art
performance on demanding datasets such as SportsMOT and Dancetrack, which
feature highly nonlinear object motion. Notably, without fine-tuning on target
datasets, MotionTrack also exhibits competitive performance on conventional
benchmarks including MOT17 and MOT20.

本文提出了一种名为 MotionTrack 的新型可学习动作预测器，其综合结合了两个层级的运动特征以增强时态动态的建模和促进个体物体准确的未来运动预测。