Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

JDAN: Joint Detection and Association Network for Real-Time Online Multi-Object Tracking

JDAN: Joint Detection and Association Network for Real-Time Online Multi-Object Tracking In the last few years, enormous strides have been made for object detection and data association, which are vital subtasks for one-stage online multi-object tracking (MOT). However, the two separated submodules involved in the whole MOT pipeline are processed or optimized separately, resulting in a complex method design and requiring manual settings. In addition, few works integrate the two subtasks into a single end-to-end network to optimize the overall task. In this study, we propose an end-to-end MOT network called joint detection and association network (JDAN) that is trained and inferred in a single network. All layers in JDAN are differentiable, and can be optimized jointly to detect targets and output an association matrix for robust multi-object tracking. What’s more, we generate suitable pseudo-labels to address the data inconsistency between object detection and association. The detection and association submodules could be optimized by the composite loss function that is derived from the detection results and the generated pseudo association labels, respectively. The proposed approach is evaluated on two MOT challenge datasets, and achieves promising performance compared with classic and latest methods. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png ACM Transactions on Multimedia Computing Communications and Applications (TOMCCAP) Association for Computing Machinery

JDAN: Joint Detection and Association Network for Real-Time Online Multi-Object Tracking

Download PDF
17 pages

Loading next page...
 
/lp/association-for-computing-machinery/jdan-joint-detection-and-association-network-for-real-time-online-2X3xmvhEQ0
Publisher
Association for Computing Machinery
Copyright
Copyright © 2023 Association for Computing Machinery.
ISSN
1551-6857
eISSN
1551-6865
DOI
10.1145/3533253
Publisher site
See Article on Publisher Site

Abstract

In the last few years, enormous strides have been made for object detection and data association, which are vital subtasks for one-stage online multi-object tracking (MOT). However, the two separated submodules involved in the whole MOT pipeline are processed or optimized separately, resulting in a complex method design and requiring manual settings. In addition, few works integrate the two subtasks into a single end-to-end network to optimize the overall task. In this study, we propose an end-to-end MOT network called joint detection and association network (JDAN) that is trained and inferred in a single network. All layers in JDAN are differentiable, and can be optimized jointly to detect targets and output an association matrix for robust multi-object tracking. What’s more, we generate suitable pseudo-labels to address the data inconsistency between object detection and association. The detection and association submodules could be optimized by the composite loss function that is derived from the detection results and the generated pseudo association labels, respectively. The proposed approach is evaluated on two MOT challenge datasets, and achieves promising performance compared with classic and latest methods.

Journal

ACM Transactions on Multimedia Computing Communications and Applications (TOMCCAP) – Association for Computing Machinery

Published: Feb 3, 2023

Keywords: Online multi-object tracking

References

  • Multitarget tracking using Siamese neural networks
    [,
  • Confidence-based data association and discriminative deep appearance learning for robust online multi-object tracking
    [,
  • Tracking without bells and whistles
    [,
  • Evaluating multiple object tracking performance: The CLEAR MOT metrics
    [,
  • Online multi-object tracking with convolutional neural networks
    [,
  • Famnet: Joint learning of feature, affinity and multi-dimensional assignment for online multiple object tracking
    [,
  • CVPR19 tracking and detection challenge: How crowded can it get? In Proceedings of CoRR (2019)
    [,
  • Pedestrian detection: A benchmark
    [,
  • A mobile vision system for robust multi-person tracking
    [,
  • Recurrent autoregressive networks for online multi-object tracking
    [,
  • Multi-level cooperative fusion of GM-PHD filters for online multiple human tracking
    [,
  • Are we ready for autonomous driving? The KITTI vision benchmark suite
    [,
  • Mask R-CNN
    [,
  • Deep residual learning for image recognition
    [,
  • Hadamard product for low-rank bilinear pooling
    [,
  • Ubernet: Training a universal convolutional neural network for low-, mid-, and high-level vision using diverse datasets and limited memory
    [,
  • The Hungarian method for the assignment problem
    [,
  • Cornernet: Detecting objects as paired keypoints
    [,
  • MOTChallenge 2015: Towards a benchmark for multi-target tracking
    [,
  • Multi-view correlation tracking with adaptive memory-improved update model
    [,
  • Learning a dynamic feature fusion tracker for object tracking
    [,
  • Robust object tracking via weight-based local sparse appearance model
    [,
  • Focal loss for dense object detection
    [,
  • Microsoft coco: Common objects in context
    [,
  • SSD: Single shot multibox detector
    [,
  • A strong baseline for multiple object tracking on vidor dataset
    [,
  • Person re-identification with expanded neighborhoods distance re-ranking
    [,
  • Multi-target tracking using CNN-based features: CNNMTT
    [,
  • MOT16: A benchmark for multi-object tracking
    [,
  • Algorithms for the assignment and transportation problems
    [,
  • Robust object tracking via local sparse appearance model
    [,
  • Multi-pattern correlation tracking
    [,
  • PyTorch: An imperative style, high-performance deep learning library
    [,
  • Hyperface: A deep multi-task learning framework for face detection, landmark localization, pose estimation, and gender recognition
    [,
  • Yolov3: An incremental improvement
    [,
  • Faster R-CNN: Towards real-time object detection with region proposal networks
    [,
  • Performance measures and a data set for multi-target, multi-camera tracking
    [,
  • Online multi-target tracking with strong and weak detections
    [,
  • Deep affinity network for multiple object tracking
    [,
  • MOTS: Multi-object tracking and segmentation
    [,
  • Non-local attention association scheme for online multi-object tracking
    [,
  • Towards real-time multi-object tracking
    [,
  • An introduction to the kalman filter
    [,
  • Simple online and realtime tracking with a deep association metric
    [,
  • Learning to track: Online multi-object tracking by decision making
    [,
  • Joint detection and identification feature learning for person search
    [,
  • Exploring image enhancement for salient object detection in low light images
    [,
  • How to train your deep multi-object tracker
    [,
  • POI: Multiple object tracking with high performance detection and appearance feature
    [,
  • Deep layer aggregation
    [,
  • Citypersons: A diverse dataset for pedestrian detection
    [,
  • Deep learning with elastic averaging SGD
    [,
  • Fairmot: On the fairness of detection and re-identification in multiple object tracking
    [,
  • Person re-identification in the wild
    [,
  • Tracking objects as points
    [,
  • Objects as points
    [,
  • Online multi-target tracking with tensor-based high-order graph matching
    [,
  • Online multi-object tracking with dual matching attention networks
    [,
  • Online multi-target tracking with tensor-based high-order graph matching
    Zhou, Zongwei; Xing, Junliang; Zhang, Mengdan; Hu, Weiming
  • Online Multi-Object Tracking with Dual Matching Attention Networks
    Zhu, Ji; Yang, Hua; Liu, Nian; Kim, Minyoung; Zhang, Wenjun; Yang, Ming-Hsuan