Multiple Object Tracking Using K-Shortest Paths Optimization

By Jérôme Berclaz, Franccois Fleuret, Engin Türetken, Pascal Fua. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011

1,095 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Management Science and Operations Research, Ocean Engineering

View Paper: https://www.semanticscholar.org/paper/a79b60109070e1da614b6fd441ff12dcffcd7908

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Management Science and Operations Research

In Ocean Engineering

Citations in the corpus

3D Pictorial Structures for Multiple Human Pose Estimation (2014, methodology): “Finally, we propose two new datasets (Campus [5] and Shelf) with ground-truth annotations and evaluate our multiple human pose estimation method.”
Learning to Track: Online Multi-object Tracking by Decision Making (2015, methodology): “In order to resolve ambiguities in associating object detections and to overcome detection failures, most of these recent works [7, 11, 27, 23] process video sequences in a batch mode in which video frames from future time steps are also utilized to solve the data association problem.”
Near-Online Multi-target Tracking with Aggregated Local Flow Descriptor (2015, background): “Due to the limited amount of observation, the tracking algorithm may produce an erroneous association at t2.”
Learning a Neural Solver for Multiple Object Tracking (2019, background): “The data association can then be formulated as maximum flow [8] or, equivalently, minimum cost problem with either fixed costs based on distance [29, 52, 73], including motion models [42], or learned costs [40].”
Simple online and realtime tracking with a deep association metric (2017, background): “For example, flow network formulations [1, 2, 3] and probabilistic graphical models [4, 5, 6, 7] have become popular frameworks of this type.”
Multiple Hypothesis Tracking Revisited (2015, methodology): “Network flow-based methods [35, 4, 45, 10] have recently become a standard approach to visual multi-target tracking due to their computational efficiency and optimality.”
Space-Time Correspondence as a Contrastive Random Walk (2020, background): “This can be addressed in a variety of ways, including dynamic programming, particle ﬁltering, various graph-based combinatorial optimization, and more recently, graph neural networks [96, 69, 6, 66, 12, 94, 38, 37, 40, 9].”
Fast YOLO: A Fast You Only Look Once System for Real-time Embedded Object Detection in Video (2017, methodology): “Second, the proposed Fast YOLO framework, O-YOLOv2, and original YOLOv2 are evaluated in terms of average run-time on a Nvidia Jetson TX1 embedded system on a video from [20].”
Joint Probabilistic Data Association Revisited (2015, background): “Within computer vision, applications of multi-target tracking are exemplified by the tasks of surveillance of a crowd of pedestrians [5, 25, 28, 29, 33, 44], and of tracking dense cellular and sub-cellular structures in biological sequences [11, 35] (Fig.”
Multiple People Tracking by Lifted Multicut and Person Re-identification (2017, background): “[3] proposes to model all potential locations over time and find trajectories that produce the minimum cost.”
Online Multi-Target Tracking Using Recurrent Neural Networks (2016, background): “Recently, a large amount of work focused on simpliﬁed models that could be solved to (near) global optimality (Jiang, Fels, and Little 2007; Zhang, Li, and Nevatia 2008; Berclaz et al. 2011; Butt and Collins 2013).”
Tracking Without Bells and Whistles (2019, background): “The data association can then be formulated as maximum flow [4] or, equivalently, minimum cost problem with either fixed costs based on distance [26, 49, 66], including motion models [39], or learned costs [36].”
Learning by Tracking: Siamese CNN for Robust Target Association (2016, methodology): “The data association can then be formulated as maximum flow [4] or, equivalently, minimum cost problem [28, 35, 45, 64], both efficiently solved to (near-)global optimality with LP, with a superior performance compared to frame-by-frame [29] or track-by-track [3] methods.”
Seq-NMS for Video Object Detection (2016, methodology): “Other classes of methods tried to compute all of the object trajectories at once using linear programming [6, 2].”
Person Re-identification: Past, Present and Future (2016, background): “This work views it as an “ultimate” goal to integrate detection, tracking, and retrieval into one framework, and evaluate the impact of each module on the [18], [77], [128], [133] for end-to-end person re-identiﬁcation (search). overall re-ID performance.”
Globally-optimal greedy algorithms for tracking a variable number of objects (2011, methodology): “describe a O(KN log N) algorithm for multi-object tracking in [4].”