An Online and Flexible Multi-object Tracking Framework Using Long Short-Term Memory

Wan, Xingyu; Wang, Jinjun; Zhou, Sanping

doi:10.1109/cvprw.2018.00169

Cited by 27 publications

(13 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They also showed that having higher-quality detections reduces the need of complex tracking algorithms while still obtaining similar results: this is because the MOTA score is heavily influenced by the amount of false positives and false negatives, and using accurate detections is an effective way of reducing both. The detections computed by [38] on the MOT16 dataset have also been made available to the public 13 and many MOT algorithms have since exploited them [41,42,43,44,45,46,47,48,49,50,51].…”

Section: Faster R-cnnmentioning

confidence: 99%

See 1 more Smart Citation

Deep learning in video multi-object tracking: A survey

et al. 2020

View full text Add to dashboard Cite

The problem of Multiple Object Tracking (MOT) consists in following the trajectory of different objects in a sequence, usually a video. In recent years, with the rise of Deep Learning, the algorithms that provide a solution to this problem have benefited from the representational power of deep models. This paper provides a comprehensive survey on works that employ Deep Learning models to solve the task of MOT on single-camera videos. Four main steps in MOT algorithms are identified, and an in-depth review of how Deep Learning was employed in each one of these stages is presented. A complete experimental comparison of the presented works on the three MOTChallenge datasets is also provided, identifying a number of similarities among the top-performing methods and presenting some possible future research directions.In this section, a general description about the problem of MOT is provided. The main characteristics and common steps of MOT algorithms are identified and described in section 2.1. The metrics that are usually employed to evaluate the performance of the models are discussed in section 2.2, while the most important benchmark datasets are presented in section 2.3. 7 The website says the detections were obtained using a model based on a latent SVM, or L-SVM. That model is now known as Deformable Parts Model (DPM). 8

show abstract

Section: Faster R-cnnmentioning

confidence: 99%

“…Wan et al [43] also used a Siamese LSTM in their algorithm, that was also composed of two steps. In the first step, short reliable tracklets were built by using Hungarian algorithm with affinity measures computed using the IoU between detections and the predicted target positions (obtained with Kalman filter or Lucas-Kanade optical flow).…”

Section: Siamese Lstmsmentioning

confidence: 99%

Deep learning in video multi-object tracking: A survey

et al. 2020

View full text Add to dashboard Cite

show abstract

“…In [19], the authors proposed dual attention matching networks (DMANs) combining spatial and temporal attention for tracking where bi‐directional LSTMs are used for modelling the attention values. Recent work by Wan et al [20] proposed a Siamese LSTM network to learn spatial and temporal trajectory components for measuring similarity for object target association. Lu et al [21] proposed a LSTM structure, namely association LSTM for regressing the object locations and labelling by learning the association features for tracked objects capturing their spatial and temporal information.…”

Section: Related Workmentioning

confidence: 99%

Multiple object tracking using feature fusion in hierarchical LSTMs

Hassan¹

2020

J. eng.

View full text Add to dashboard Cite

“…Milan et al [33] presented an on-line RNN-based approach for multiple people tracking which is capable of performing prediction, data association and state update within a unified network structure. Followed by these works, [56] extended the research of RNN-based methods and leveraged the power of Long Short-Term Memory (LSTM) for learning a discriminative model of object trajectory by integrating dynamic features both in temporal and spatial. For the on-line MOT task, these methods may not perform well when heavy occlusion or mis-detection downgrade the robustness of appearance model.…”

Section: Related Workmentioning

confidence: 99%

“…Furthermore, aiming to learn a robust metric for feature representation, several works [37,47,45] take multiple features of objects in the scene by incorporating a myriad of components such as motion, appearance, interaction, etc. Some works [56,24] even consider to combine temporal components to analyze long-term variation by using Long Short-Term Memory (LSTM). Since these methods are still based on disjoint detection/association steps, the computation is huge, and the performance is limited without end-to-end (i.e., from image-sequence/video to trajectory) capacity.…”

Section: Introductionmentioning

confidence: 99%

End-to-End Multi-Object Tracking with Global Response Map

Wan¹,

Cao²,

Zhou³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Most existing Multi-Object Tracking (MOT) approaches follow the Tracking-by-Detection paradigm and the data association framework where objects are firstly detected and then associated. Although deep-learning based method can noticeably improve the object detection performance and also provide good appearance features for cross-frame association, the framework is not completely end-to-end, and therefore the computation is huge while the performance is limited. To address the problem, we present a completely end-to-end approach that takes imagesequence/video as input and outputs directly the located and tracked objects of learned types. Specifically, with our introduced multi-object representation strategy, a global response map can be accurately generated over frames, from which the trajectory of each tracked object can be easily picked up, just like how a detector inputs an image and outputs the bounding boxes of each detected object. The proposed model is fast and accurate. Experimental results based on the MOT16 and MOT17 benchmarks show that our proposed on-line tracker achieved state-of-the-art performance on several tracking metrics.

show abstract

An Online and Flexible Multi-object Tracking Framework Using Long Short-Term Memory

Cited by 27 publications

References 28 publications

Deep learning in video multi-object tracking: A survey

Deep learning in video multi-object tracking: A survey

Multiple object tracking using feature fusion in hierarchical LSTMs

End-to-End Multi-Object Tracking with Global Response Map

Contact Info

Product

Resources

About