Simple online and realtime tracking

By Alex Bewley, ZongYuan Ge, Lionel Ott, Fabio Tozeto Ramos, Ben Upcroft. 2016

4,063 citations recorded by the source corpus.

This paper explores a pragmatic approach to multiple object tracking where the main focus is to associate objects efficiently for online and realtime applications. To this end, detection quality is identified as a key factor influencing tracking performance, where changing the detector can improve tracking by up to 18.9%. Despite only using a rudimentary combination of familiar techniques such as the Kalman Filter and Hungarian algorithm for the tracking components, this approach achieves an accuracy comparable to state-of-the-art online trackers. Furthermore, due to the simplicity of our tracking method, the tracker updates at a rate of 260 Hz which is over 20x faster than other state-of-the-art trackers.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

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

In Computer Vision and Pattern Recognition

Citations in the corpus

Joint Object Detection and Multi-Object Tracking with Graph Neural Networks (2021, methodology): “Online methods [10], [14], [44], [45] only require information up to the current frame for data association and can be useful to online applications.”
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]).”
Track to Detect and Segment: An Online Multi-Object Tracker (2021, methodology): “MOT was dominated by the tracking-by-detection ( TBD ) paradigm over the past years [58, 6, 66, 52, 33, 5, 38, 48, 54].”
Quasi-Dense Similarity Learning for Multiple Object Tracking (2020, methodology): “Some methods use motion priors, such as Kalman Filter [4, 52], optical flow [48], and displacement regression [10, 16], to ensure accurate distance estimations.”
TransTrack: Multiple-Object Tracking with Transformer (2020, methodology): “SORT [4] tracks bounding boxes using a Kalman Filter [40] and associates each bounding box with its highest overlapping detection in the current frame using the Hungarian algorithm.”
Joint Hand Motion and Interaction Hotspots Prediction from Egocentric Videos (2022, background): “Kalman Filter (KF) [7] tracks the center of the hand in observation frames and predicts future hands locations.”
Segment as Points for Efficient Online Multi-Object Tracking and Segmentation (2020, positive): “DeepSort extends SORT [3] by incorporating”
Next-active-object prediction from egocentric videos (2017, methodology): “This can be done using a tracker based on a data association algorithm such as the real-time (260 Hz) SORT tracker proposed in [46].”
Joint Monocular 3D Vehicle Detection and Tracking (2018, methodology): “We compare with Sort [4], Deep Sort [46] tracker for inter-frame ROIs matching in 2D.”
MOTR: End-to-End Multiple-Object Tracking with TRansformer (2021, methodology): “DeepSORT [39] and Tracktor [2] introduced an extra cosine distance and compute the appearance similarity for track association.”
An Edge Traffic Flow Detection Scheme Based on Deep Learning in an Intelligent Transportation System (2021, positive): “DeepSORT integrates appearance information to improve the performance of SORT [41] by adding one pretrained association metric.”
Deep OC-Sort: Multi-Pedestrian Tracking by Adaptive Re-Identification (2023, methodology): “Among this line of work, SORT [9] uses a linear motion assumption to associate tracks by IoU.”
Egocentric Hand Track and Object-Based Human Action Recognition (2019, methodology): “Hand bounding boxes are associated through the course of frames and identified as belonging to a track with a certain id. SORT uses the Kalman filter [32] to predict the coordinates of what would likely be the next bounding box of an existing track and the Hungarian algorithm [33] to assign the dete”
ByteTrack: Multi-Object Tracking by Associating Every Detection Box (2021, methodology): “We compare BYTE and DeepSORT [22] using light detection models.”
Opening up Open World Tracking (2021, methodology): “As can be seen, OWTB performs signiﬁcantly better compared to SORT in terms of detection recall ( +9 .”