Detection confidence driven multi-object tracking to recover reliable tracks from unreliable detections

Mandel, Travis; Jimenez, Mark; Risley, Emily; Nammoto, Taishi; Williams, Rebekka; Panoff, Max; Ballesteros, Meynard; Suarez, Bobbie

doi:10.1016/j.patcog.2022.109107

Cited by 15 publications

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[1,30,37,58,62]). In [39] authors proposed an offline tracking algorithm that uses all detection boxes. LG-Track [40] uses localization and classification confidence scores from the detectors and divides detections into four groups based on thresholds for the two scores.…”

Section: Working With Unreliable Detectionsmentioning

confidence: 99%

BoostTrack: boosting the similarity measure and detection confidence for improved multiple object tracking

Stanojevic,

Todorovic

2024

Machine Vision and Applications

View full text Add to dashboard Cite

Handling unreliable detections and avoiding identity switches are crucial for the success of multiple object tracking (MOT). Ideally, MOT algorithm should use true positive detections only, work in real-time and produce no identity switches. To approach the described ideal solution, we present the BoostTrack, a simple yet effective tracing-by-detection MOT method that utilizes several lightweight plug and play additions to improve MOT performance. We design a detection-tracklet confidence score and use it to scale the similarity measure and implicitly favour high detection confidence and high tracklet confidence pairs in one-stage association. To reduce the ambiguity arising from using intersection over union (IoU), we propose a novel Mahalanobis distance and shape similarity additions to boost the overall similarity measure. To utilize low-detection score bounding boxes in one-stage association, we propose to boost the confidence scores of two groups of detections: the detections we assume to correspond to the existing tracked object, and the detections we assume to correspond to a previously undetected object. The proposed additions are orthogonal to the existing approaches, and we combine them with interpolation and camera motion compensation to achieve results comparable to the standard benchmark solutions while retaining real-time execution speed. When combined with appearance similarity, our method outperforms all standard benchmark solutions on MOT17 and MOT20 datasets. It ranks first among online methods in HOTA metric in the MOT Challenge on MOT17 and MOT20 test sets. We make our code available at https://github.com/vukasin-stanojevic/BoostTrack.

show abstract

Section: Working With Unreliable Detectionsmentioning

confidence: 99%

BoostTrack: boosting the similarity measure and detection confidence for improved multiple object tracking

Stanojevic,

Todorovic

2024

Machine Vision and Applications

View full text Add to dashboard Cite

show abstract

“…With the advancements in deep learning and computer vision in recent years, computer vision based methods are increasingly being employed in the study of fish behavior (Li et al, 2020;Yang et al, 2020;. Fish multiple object tracking technology has demonstrated increasingly powerful capabilities, efficiently and accurately capturing fish trajectories (Li, W. et al, 2022;Mandel et al, 2023;Liu et al, 2024). Compared to manual and acoustic-based methods, computer vision-based fish multiple object tracking approaches offer lower costs and higher efficiency, enabling rapid processing of large volumes of videos to obtain fish behavior information for further study.…”

Section: Introductionmentioning

confidence: 99%

Behavior analysis of juvenile steelhead trout under blue and red light color conditions based on multiple object tracking

Li,

Chen,

Huang

et al. 2024

Front. Mar. Sci.

View full text Add to dashboard Cite

IntroductionThe lighting environment significantly influences fish behavior. This study explores the impact of diverse lighting conditions on the behavior of steelhead trout (Oncorhynchus mykiss) to illuminate the mechanisms underlying their behavioral responses.MethodsThis experiment was set up with six treatments at a constant light intensity of 150 lx: 12h white light + 12h dark (12 W), 12h blue light + 12h dark (12B), 12h red light + 12h dark (12 R), 1.5h blue light + 9h red light + 1.5h blue light + 12h dark (3B9R), 3h blue light + 6h red light + 3h blue light + 12h dark (6B6R), total 12h of blue and red light + 12h dark (T12BR). A multiple object tracking method, YOLOv5 with SORT, was employed to capture the movement trajectory of each fish, quantifying three motion metrics: swimming velocity, swimming angular velocity, and generalized intersection over union.ResultsThe results revealed that fish exposed to 12R light environment showed significantly higher activity levels than other groups. The mixed light environments (3B9R, 6B6R) formed significant differences in behavioral metrics with 12R earlier than pure light environments (12B, 12W, T12BR), indicating sudden light color changes should be avoided. Fish in the 3B9R environment exhibited the lowest activity level but highest growth performance, with the highest specific growth rate of 1.91±0.12 d-1, a value significantly surpassing the lowest recorded rate, supported by a p-value of 0.0054, indicating it is suitable for steelhead trout cultivation.DiscussBehavioral significant differences were observed as early as week eight, much earlier than physiological differences, which became apparent by week 16. Overall, this paper employs computer vision methods to study the impact of different light colors on fish behavior, found that 3B9R is the optimal lighting condition tested and sudden light color changes should be avoided, offering a new perspective on light conditions and behavior in steelhead trout cultivation.

show abstract

“…It incorporates an iterative graph clustering strategy for efficient proposal generation and employs a trainable Graph-Convolutional Network (GCN) for accurate proposal scoring, capturing higher-order information within each proposal. Targeting MOT tasks in challenging environment, Mandel et al [24] proposed the RCT algorithm, which enhances the tracking quality throughout the entire tracking process based on precise detection confidence values, and achieves excellent performance in real-world underwater fish tracking. This paper focuses on online multi-object tracking, which is more valuable in real-world scenarios.…”

Section: Introductionmentioning

confidence: 99%

A deep learning framework for multi‐object tracking in team sports videos

Cao,

Wang,

Liu

et al. 2024

IET Computer Vision

View full text Add to dashboard Cite

In response to the challenges of Multi‐Object Tracking (MOT) in sports scenes, such as severe occlusions, similar appearances, drastic pose changes, and complex motion patterns, a deep‐learning framework CTGMOT (CNN‐Transformer‐GNN‐based MOT) specifically for multiple athlete tracking in sports videos that performs joint modelling of detection, appearance and motion features is proposed. Firstly, a detection network that combines Convolutional Neural Networks (CNN) and Transformers is constructed to extract both local and global features from images. The fusion of appearance and motion features is achieved through a design of parallel dual‐branch decoders. Secondly, graph models are built using Graph Neural Networks (GNN) to accurately capture the spatio‐temporal correlations between object and trajectory features from inter‐frame and intra‐frame associations. Experimental results on the public sports tracking dataset SportsMOT show that the proposed framework outperforms other state‐of‐the‐art methods for MOT in complex sport scenes. In addition, the proposed framework shows excellent generality on benchmark datasets MOT17 and MOT20.

show abstract

Detection confidence driven multi-object tracking to recover reliable tracks from unreliable detections

Cited by 15 publications

References 16 publications

BoostTrack: boosting the similarity measure and detection confidence for improved multiple object tracking

BoostTrack: boosting the similarity measure and detection confidence for improved multiple object tracking

Behavior analysis of juvenile steelhead trout under blue and red light color conditions based on multiple object tracking

A deep learning framework for multi‐object tracking in team sports videos

Contact Info

Product

Resources

About