Regional ViT-adaptive dilated YOLOv10 for traffic sign detection with variational conditional DDGAN-based image dehazing

Automatically detecting and recognizing traffic sign patterns are important for managing traffic-sign inventory systems. Accurate recognition and detection of traffic sign patterns remain a widely researched challenge in the computer vision domain. Nevertheless, existing approaches focus on only a comparatively small subset of traffic signs, while effective detection across the broader range of traffic sign categories, including small, distant, partially occluded, and low-visibility signs under adverse weather conditions, is essential for fully automated traffic-sign inventory management, thereby reducing the need for manual processes. Furthermore, detecting and recognizing small traffic signs with higher accuracy in real-time scenarios is crucial. Therefore, this study introduces an innovative traffic sign detection framework to improve detection precision, particularly for small-scale and low-resolution distant regions. Initially, the related images for this proposed study are gathered from the available datasets. Subsequently, these images undergo an image dehazing process using the Variational Conditional Dual Discriminator Generative Adversarial Network (VDDGAN) model. The enhanced images are provided to the Regional Vision Transformer-Adaptive Dilated You Only Look Oncev10 (RViT-ADYv10) for traffic sign detection. In this framework, the parameters of YOLOv10 are optimally tuned using the Predatory Search-based Frilled Lizard Optimization (PSFLO) to improve detection performance. Finally, the model's effectiveness is analyzed with traditional approaches to validate the robustness of the designed framework. When considering the BDD100K, TT100K, and GTSDB datasets, our PSFLO-RViT-ADYv10 model achieved [email protected] scores of 89.8%, 87.9%, and 86.7%, respectively, with maximum sensitivities of 94.83%, 93.9%, and 95.83%, which ensures robust and efficient traffic sign detection performance across diverse driving environments.

Authors

Institutions

Publication Details

Journal
Frontiers in Computer Science
Published
2026-11-17
DOI
https://doi.org/10.3389/fcomp.2026.1903292
Primary Topic
Advanced Neural Network Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Regional ViT-adaptive dilated YOLOv10 for traffic sign detection with variational conditional DDGAN-based image dehazing

C M Vidhyapathi, Thiyagarajan V
1 links
Frontiers in Computer Science
Advanced Neural Network Applications
article

Regional ViT-adaptive dilated YOLOv10 for traffic sign detection with variational conditional DDGAN-based image dehazing

C M Vidhyapathi, Thiyagarajan V
article en

Abstract

Automatically detecting and recognizing traffic sign patterns are important for managing traffic-sign inventory systems. Accurate recognition and detection of traffic sign patterns remain a widely researched challenge in the computer vision domain. Nevertheless, existing approaches focus on only a comparatively small subset of traffic signs, while effective detection across the broader range of traffic sign categories, including small, distant, partially occluded, and low-visibility signs under adverse weather conditions, is essential for fully automated traffic-sign inventory management, thereby reducing the need for manual processes. Furthermore, detecting and recognizing small traffic signs with higher accuracy in real-time scenarios is crucial. Therefore, this study introduces an innovative traffic sign detection framework to improve detection precision, particularly for small-scale and low-resolution distant regions. Initially, the related images for this proposed study are gathered from the available datasets. Subsequently, these images undergo an image dehazing process using the Variational Conditional Dual Discriminator Generative Adversarial Network (VDDGAN) model. The enhanced images are provided to the Regional Vision Transformer-Adaptive Dilated You Only Look Oncev10 (RViT-ADYv10) for traffic sign detection. In this framework, the parameters of YOLOv10 are optimally tuned using the Predatory Search-based Frilled Lizard Optimization (PSFLO) to improve detection performance. Finally, the model's effectiveness is analyzed with traditional approaches to validate the robustness of the designed framework. When considering the BDD100K, TT100K, and GTSDB datasets, our PSFLO-RViT-ADYv10 model achieved [email protected] scores of 89.8%, 87.9%, and 86.7%, respectively, with maximum sensitivities of 94.83%, 93.9%, and 95.83%, which ensures robust and efficient traffic sign detection performance across diverse driving environments.

Frontiers in Computer ScienceVol. 8
Vellore Institute of Technology University (IN)
Sustainable cities and communities, Industry, innovation and infrastructure
Advanced Neural Network Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon