Learning to Track: Online Multi-object Tracking by Decision Making

By Yu Xiang, Alexandre Alahi, Silvio Savarese. IEEE International Conference on Computer Vision, 2015

679 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Global and Planetary Change

View Paper: https://www.semanticscholar.org/paper/3227b933d917c73902746ac2c36a72927064a3c0

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

In Global and Planetary Change

Citations in the corpus

Joint Graph Decomposition & Node Labeling: Problem, Algorithms, Applications (2016, background): “ns deﬁne both a decomposition and a node labeling of a givengraph(Fig. 1). This problem offers a common mathematical abstraction of seemingly unrelated computer vision tasks: Multiple object tracking [2, 3, 4, 7, 13, 17, 19, 29, 30] can be seen as the task of deciding, for every point in an image, w”
Learning Social Etiquette: Human Trajectory Understanding In Crowded Scenes (2016, methodology): “We evaluate our proposed MTT algorithm against the following baselines: (i) Xiang's MDP algorithm [39] with a linear motion prior, (ii) [39] with single class forecasting model [1], (iii) [39] with our proposed multi-class forecasting model based on social sensitivity.”
Virtual KITTI 2 (2020, neutral): “[17].”
Joint Monocular 3D Vehicle Detection and Tracking (2018, methodology): “[47] enhance tracking by modeling a track-let into different state and explicitly learns a MDP for state transition.”
Multi-Scale Deep Reinforcement Learning for Real-Time 3D-Landmark Detection in CT Scans (2019, methodology): “Reinforcement learning is also used to model the lifetime and visibility of objects in online tracking [63].”
Deep Network Flow for Multi-object Tracking (2017, background): “ribe the path of individual object instances over time. Bipartite graph matching [25,35] is often employed in on-line approaches to assign bounding boxes in the current frame to existing trajectories [22,37,38,52]. Off-line methods can be elegantly formulated in a network ﬂow framework to solve the ”
Online Multi-Object Tracking with Dual Matching Attention Networks (2018, methodology): “In contrast, online MOT methods [53,10,27] do not rely on detections from future frames and may not perform well when target objects are heavily occluded or mis-detected.”
Multiple People Tracking by Lifted Multicut and Person Re-identification (2017, background): “The spatio-temporal relation based feature is commonly used in many multi-person tracking works [23, 34, 6], as it is a good affinity measure for pairs of detections that are in close proximity.”
The Unmanned Aerial Vehicle Benchmark: Object Detection and Tracking (2018, methodology): “Therefore, we evaluate 8 recent algorithms including online methods (CMOT [2], MDP [50], SORT [6] and DSORT [48]) and batch based methods (GOG [35], CEM [30], SMOT [13] and IOUT [7]).”
Deep learning in video multi-object tracking: A survey (2019, methodology): “The two probabilities, that can be easily seen as a kind of affinity measure, were then multiplied together to obtain a final association probability, used to solve a bipartite matching problem for association between tracklets and detections following the algorithm in [134].”
Computer Vision for Autonomous Vehicles: Problems, Datasets and State-of-the-Art (2017, methodology): “In contrast to the Bayesian and min-cost formulation, Xiang et al. (2015a) consider the tracking problem as a Markov decision process (MDP).”
Simple online and realtime tracking (2016, methodology): “For brevity, only the most relevant trackers, which are state-of-the-art online trackers in terms of accuracy, such as ( TDAM [18], MDP [12]), the fastest batch based tracker ( DP NMS [23]), and all round near online method ( NOMT [11]) are listed.”
VirtualWorlds as Proxy for Multi-object Tracking Analysis (2016, methodology): “Thanks to the recent progress on object detection, association-based tracking-by-detection in monocular video streams is particularly successful and widely used for MOT [29, 30, 31, 32, 33, 34, 35, 36, 37, 38] (see [39] for a recent review).”
GSM: Graph Similarity Model for Multi-Object Tracking (2020, background): “Many early works extracted hand-craft appearance feature such as raw pixel template [ Yamaguchi et al. , 2011; Pellegrini et al. , 2009 ] , color histogram [ Izadinia et al. , 2012; Xiang et al. , 2015 ] , HOG [ Izadinia et al. , 2012; Kuo et al. , 2010 ] and so on.”
Online Multi-object Tracking Using CNN-Based Single Object Tracker with Spatial-Temporal Attention Mechanism (2017, methodology): “Some previous works [51, 53, 5, 52, 59, 50] have attempted to adopt single object tracking methods into MOT problem.”
Video Instance Segmentation Tracking With a Modified VAE Architecture (2020, methodology): “Detection-based tracking [27, 67, 63, 20, 49, 13, 4, 61, 32, 42], on the other hand, generally requires objects being detected followed by a tracker that links the detection regions to form trajectories of the targets.”
Spatial-Temporal Relation Networks for Multi-Object Tracking (2019, methodology): “Most recent approaches follow the popular “tracking-by-detection” paradigm [12, 19, 27, 33, 35, 47, 58], where objects are firstly localized in each frame and then associated across frames.”
Tracking Objects as Points (2020, background): “s currently the dominant tracking paradigm. Yet the best-performing object trackers are not without drawbacks. Many rely on slow and complex association strategies to link detected boxes through time [14,41,48,50]. Recent work on simultaneous detection and tracking [1,8] has made progress in allevia”
Online Multi-Target Tracking Using Recurrent Neural Networks (2016, background): “Overall, our approach does not quite reach the top accuracy in pedestrian online tracking (Xiang, Alahi, and Savarese 2015), but is two orders of magnitude faster.”
MOTChallenge: A Benchmark for Single-Camera Multiple Target Tracking (2020, background): “With the combination of the powerful MHT tracking framework (Reid 1979) and online-adapted features used for data association, this method surpasses MDP and attains over 32% MOTA on MOT15 and 45% MOTA on MOT16 .”