Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 41 of the 41 references it cites were found in a registry or an index; 2 of the 8 sources its claims cite were checked for support.

Realtime Vehicle Tracking Method Based on YOLOv5 + DeepSORT

Open access
Lin, Lixiong, He, Hongqin, Xu, Zhiping, Wu, Dongjie · 2023 · Computational Intelligence and Neuroscience

Abstract

In actual traffic scenarios, the environment is complex and constantly changing, with many vehicles that have substantial similarities, posing significant challenges to vehicle tracking research based on deep learning. To address these challenges, this article investigates the application of the DeepSORT (simple online and realtime tracking with a deep association metric) multitarget tracking algorithm in vehicle tracking. Due to the strong dependence of the DeepSORT algorithm on target detection, a YOLOv5s_DSC vehicle detection algorithm based on the YOLOv5s algorithm is proposed, which provides accurate and fast vehicle detection data to the DeepSORT algorithm. Compared to YOLOv5s, YOLOv5s_DSC has no more than a 1% difference in optimal mAP0.5 (mean average precision), precision rate, and recall rate, while reducing the number of parameters by 23.5%, the amount of computation by 32.3%, the size of the weight file by 20%, and increasing the average processing speed of each image by 18.8%. After integrating the DeepSORT algorithm, the processing speed of YOLOv5s_DSC + DeepSORT reaches up to 25 FPS, and the system exhibits better robustness to occlusion.

Overview

CoChat Summary

TL;DR

The paper proposes a vehicle tracking pipeline that combines an improved YOLOv5s detector (YOLOv5s_DSC) with DeepSORT to improve realtime performance and robustness in occluded traffic scenes. On their reported test setup, the combined system reaches 25 FPS and the detector keeps mAP0.5, precision, and recall within 1% of YOLOv5s while reducing parameters, FLOPS, and model size.

Key findings

  • YOLOv5s_DSC is reported to differ from YOLOv5s by no more than 1% in optimal mAP0.5, precision, and recall, while reducing parameters by 23.5%, computation by 32.3%, and weight-file size by 20%; average image processing speed increases by 18.8%.
  • In Table 5, YOLOv5s reports mAP0.5 = 0.993, mAP0.5:0.95 = 0.865, accuracy = 0.981, recall = 0.983, parameters = 7,071,633, computation = 16.4 GFLOPS, and model size = 14.4 M; YOLOv5s_DSC reports mAP0.5 = 0.993, mAP0.5:0.95 = 0.851, accuracy = 0.980, recall = 0.976, parameters = 5,622,481, computation = 12.3 GFLOPS, and model size = 11.5 M.
  • After integrating DeepSORT, the YOLOv5s_DSC + DeepSORT system reaches up to 25 FPS and is described as more robust to occlusion.
  • The DeepSORT appearance network trained on the vehicle re-recognition dataset reaches 88% test accuracy after 100 epochs.

Methods

  • The study proposes a two-stage vehicle tracking framework: YOLOv5s_DSC for vehicle detection followed by DeepSORT for multi-object tracking, using Kalman filtering and Hungarian matching with Mahalanobis distance plus cosine appearance distance.
  • Experiments use the VeRi dataset for training the DeepSORT re-identification network and the UA-DETRAC dataset for training/evaluating YOLOv5s and YOLOv5s_DSC; the VeRi split is 8:1 and UA-DETRAC is split 9:1, with 73,876 training and 8,209 test images reported for UA-DETRAC.
  • The main analysis compares YOLOv5s and YOLOv5s_DSC on mAP, precision, recall, parameter count, GFLOPS, model size, and image processing speed, and then verifies tracking on traffic videos with frontal and oblique views plus occlusion scenarios.

Limitations

  • The authors state that the algorithm does not consider detection performance in adverse weather or blurred vehicle videos such as rainy days and foggy weather.
  • The paper notes data used to support the findings are available upon request from the corresponding author, which implies the underlying data are not openly provided in the text.
  • Evidence from the text: the evaluation is limited to the reported datasets and traffic videos described in the paper, so generalization beyond those scenarios is not demonstrated here.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1155/2023/7974201

Cites

41 of 41 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Block diagram of the DeepSORT algorithm.
Figure 1 Block diagram of the DeepSORT algorithm.CC-BY · From the PDF
Figure 2 Depthwise convolution.
Figure 2 Depthwise convolution.CC-BY · From the PDF
Figure 4 Te frst C3 structure of YOLOv5s.
Figure 4 Te frst C3 structure of YOLOv5s.CC-BY · From the PDF
Figure 5 DSC structure with residual.
Figure 5 DSC structure with residual.CC-BY · From the PDF
Image 5
CC-BY · From the PDF
Figure 8 Training loss graph of DeepSORT deep convolutional neural network.
Figure 8 Training loss graph of DeepSORT deep convolutional neural network.CC-BY · From the PDF
Figure 9 Comparison of training loss between YOLOv5s and YOLOv5s_DSC.
Figure 9 Comparison of training loss between YOLOv5s and YOLOv5s_DSC.CC-BY · From the PDF
Figure 10 mAP, accuracy, and recall of YOLOv5s and YOLOv5s_DSC.
Figure 10 mAP, accuracy, and recall of YOLOv5s and YOLOv5s_DSC.CC-BY · From the PDF
Image 9
CC-BY · From the PDF
Figure 11 YOLOv5s_DSC vehicle detection at the frontal windows.
Figure 11 YOLOv5s_DSC vehicle detection at the frontal windows.CC-BY · From the PDF
Figure 13 YOLOv5s_DSC vehicle detection efect in front of an intersection.
Figure 13 YOLOv5s_DSC vehicle detection efect in front of an intersection.CC-BY · From the PDF
Figure 14 YOLOv5s_DSC+DeepSORT recognition and tracking efect.
Figure 14 YOLOv5s_DSC+DeepSORT recognition and tracking efect.CC-BY · From the PDF
Image 13
CC-BY · From the PDF

Verification

Claims the paper makes

1 Valid · 1 Not valid · 5 Not checked · 42 Own results and interpretations

19 more claims

Create a free account to see all 49 claims and to verify the 1 reference still open.

See all claims — free account
  • It may state two separate claims as one. 22
  • Probably not a substantial claim. 4
  • The quote only partly states this claim. 3

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Self-check results appear here once an independent check has confirmed them.

Cited works that were retracted

Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.

1 of 1

Citations

Cites

1 retracted

Show all 38 references

Research works printed but not linked: 2

  1. Yan G., 2017, Real-time L1-tracker based on hear-like features, Computer Science, 44, 300
  2. Li J., 2019, YOLOv3 based object tracking method, Electronics Optics and Control, 26

Cited by

See all 18

Metadata

Authors
Lin, L., He, H., Xu, Z., & Wu, D.
Affiliations
  • Jimei University (CN)
  • Dahua Technology (China) (CN)
  • Xiamen University (CN)
Corresponding author
Lin, Lixiong
Venue
Computational Intelligence and Neuroscience
Publisher
Wiley
Published
2023-01
Volume
2023
Issue
1
ISSN
  • 1687-5265 (ISSN-L)
  • 1687-5273
OpenAlex
W4380882674
Cited by
18
References
38
Citations by year
2026: 4 · 2025: 6 · 2024: 7 · 2023: 1
FWCI
1.64
Type
journal article
Language
en
Open access
Locations
Computational Intelligence and Neuroscience (publishedVersion, cc-by)
Funders
  • Jimei University (ZQ2022002)
  • Department of Education, Fujian Province (JAT220169)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Video Surveillance and Tracking Methods
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Engineering › Automotive Engineering › Autonomous Vehicle Technology and Safety
Keywords
vehicle tracking, multi-target tracking, DeepSORT, YOLOv5, vehicle detection, real-time tracking, occlusion robustness, lightweight models, traffic scenarios

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Realtime Vehicle Tracking Method Based on YOLOv5 + DeepSORT. https://app.cochat.ai/papers/10.1155/2023/7974201
The paper
Lin, L., He, H., Xu, Z., & Wu, D. (2023). Realtime Vehicle Tracking Method Based on YOLOv5 + DeepSORT. Computational Intelligence and Neuroscience. https://doi.org/10.1155/2023/7974201

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.