Global data association for multi-object tracking using network flows

By Li Zhang, Yuan Li, Ramkant Nevatia. 2008 IEEE Conference on Computer Vision and Pattern Recognition, 2008

1,090 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/6b866ba1dfd2a7b3bbae2518667c8244a6b79aaa

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 Computer Vision and Pattern Recognition

Citations in the corpus

Towards Real-Time Multi-Object Tracking (2019, background): “Data association thus can be solved by minimizing some fixed [15,26,44] or learned [19] cost, or by more complex optimization such as multi-cuts [35] and minimum cliques [43].”
A Unified Framework for Multi-target Tracking and Collective Activity Recognition (2012, background): “Note that the number of nodes in [15, 16] is usually in the order of tens or hundreds of thousands.”
Deep Network Flow for Multi-object Tracking (2017, background): “An elegant solution to assign trajectories to detections is the network ﬂow formulation [54] (see Section 3.1 for details).”
You'll never walk alone: Modeling social behavior for multi-target tracking (2009, background): “To this end, researchers have focused on improving the appearance model [10, 5], the object detector [2, 7, 9, 22], and/or the optimization strategy [14, 16, 23].”
Multiobject tracking as maximum weight independent set (2011, methodology): “Global optimal trajectory association has also been formulated as the min-cost flow problem in [24].”
An Empirical Study of Detection-Based Video Instance Segmentation (2019, methodology): “The top-performing class of tracking-bydetection methods [21, 28] link individual detections from each frame into object tracks via association.”
DETRAC: A New Benchmark and Protocol for Multi-Object Tracking (2015, background): “To date, most effective MOT approaches, e.g., [41, 6, 7, 24, 29, 39, 22, 37, 11], follow the trackingby-detection paradigm, which entails an object detection step to find target locations in each video frame (usually in the form of their bounding boxes), and a target association step that generates…”
MOTChallenge 2015: Towards a Benchmark for Multi-Target Tracking (2015, neutral): “Since its original publication [57], a large number of methods that are based on the network flow formulation have appeared in the literature [13], [31], [34], [41], [50].”
On pairwise costs for network flow multi-object tracking (2014, methodology): “While earlier min-cost network flow optimization methods have used linear programming, recently proposed solutions to the mincost flow optimization include push-relabel methods [29], successive shortest paths [22, 5], and dynamic programming [22].”
Learning to associate: HybridBoosted multi-target tracker for crowded scene (2009, methodology): “To the best of our knowledge, there has been no extensive use of machine learning algorithm in building the affinity model, except for computing simple statistics such as the mean and variance of appearance difference and detection rate [6][8].”
Multiple People Tracking by Lifted Multicut and Person Re-identification (2017, background): “Recent works on multi-person tracking focus on the tracking-by-detection approach [21, 36, 35, 29, 30].”
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].”
Discrete-continuous optimization for multi-target tracking (2012, background): “However, the continuous aspect of trajectory estimation suf-To fers, either because trajectories have to be pre-computed in absence of any data association [26, 27], or the trajectories are spatially discretized [2, 4].”
FairMOT: On the Fairness of Detection and Re-identification in Multiple Object Tracking (2020, methodology): “[28] build a graphical model with nodes representing detections in all frames for multi-object tracking.”
Multi-scale object candidates for generic object tracking in street scenes (2016, methodology): “We compare to two state-of-the-art tracking-by-detection methods [2], [5], for which we obtained tracking results from the authors.”
TrackFormer: Multi-Object Tracking with Transformers (2021, methodology): “Graphs have been used for track association and longterm re-identification by formulating the problem as a maximum flow (minimum cost) optimization [3] with distance based [20, 36, 62] or learned costs [24].”
Recent advances and trends in visual tracking: A review (2011, methodology): “In Zhang et al. [101], proposed a network ﬂow based optimization method for data association needed for multiple objects tracking.”
Multi-cue onboard pedestrian detection (2009, methodology): “We also outperform Zhang et al. [ 36 ] who report 64.3% recall at 1.5 FFPI even though their detector is trained on ETH-02 and ETH-03 whereas our detector is trained on an independent and more general multi-view training set.”
Improving Data Association by Joint Modeling of Pedestrian Trajectories and Groupings (2010, methodology): “The method however does not need the entire time window either, as global approaches [17, 20].”