Tracking Objects as Pixel-wise Distributions

Zelin Zhao, Ze Wu, Yueqing Zhuang, Boxun Li, Jiaya Jia
7/12/2022
cs.CVcs.AIcs.LG

No official code found. Best implementation: JIA-Lab-research/ECCV22-P3AFormer-Tracking-Objects-as-Pixel-wise-Distributions (159 stars, last commit 2022-09-21).

Abstract

Multi-object tracking (MOT) requires detecting and associating objects through frames. Unlike tracking via detected bounding boxes or tracking objects as points, we propose tracking objects as pixel-wise distributions. We instantiate this idea on a transformer-based architecture, P3AFormer, with pixel-wise propagation, prediction, and association. P3AFormer propagates pixel-wise features guided by flow information to pass messages between frames. Furthermore, P3AFormer adopts a meta-architecture to produce multi-scale object feature maps. During inference, a pixel-wise association procedure is proposed to recover object connections through frames based on the pixel-wise prediction. P3AFormer yields 81.2\% in terms of MOTA on the MOT17 benchmark -- the first among all transformer networks to reach 80\% MOTA in literature. P3AFormer also outperforms state-of-the-arts on the MOT20 and KITTI benchmarks.

View on arXivView PDF

Code Implementations(1)

JIA-Lab-research/ECCV22-P3AFormer-Tracking-Objects-as-Pixel-wise-DistributionsCommunity100%

The official code for our ECCV22 oral paper: tracking objects as pixel-wise distributions.

1594C++, CudaJul 15, 20224 years agoNOASSERTION
motmultiple-object-trackingp3aformertrackingtransformer

Similar Papers

Leveraging the Power of Data Augmentation for Transformer-based Tracking

87% match1 repo142023

Correlation-Embedded Transformer Tracking: A Single-Branch Framework

85% match2 repos152024

FutrTrack: A Camera-LiDAR Fusion Transformer for 3D Multiple Object Tracking

83% match2025

TransFiner: A Full-Scale Refinement Approach for Multiple Object Tracking

82% match2022

Adaptively Bypassing Vision Transformer Blocks for Efficient Visual Tracking

82% match2024

Lattice Boltzmann Model for Learning Real-World Pixel Dynamicity

82% match2025

Cite this paper

@article{zhao2022tracking,
  title  = {Tracking Objects as Pixel-wise Distributions},
  author = {Zelin Zhao and Ze Wu and Yueqing Zhuang and Boxun Li and Jiaya Jia},
  year   = {2022},
  eprint = {2207.05518},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2207.05518v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2207.05518.svg)](https://codeofpaper.com/paper/2207.05518)

Discussion