Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

OM-YOLO: A Robust and Efficient Detection Model for Preserving Features on Small Traffic Signs

Rektorat UDINUS 2026 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2026
Goals
Record
f585c420-bf15-4127-9974-bec71e06553a
Published
2026-10-30

Universitas Dian Nuswantoro (UDINUS), through researchers Pradana, Afu Ichsan; Rustad, Supriadi; Shidik, Guruh Fajar; and Santoso, Heru Agus, conducted a study on OM-YOLO, a robust and efficient detection model for preserving features on small traffic signs. The activity addressed the difficulty of detecting small-scale traffic signs in intelligent transportation systems, where spatial resolution degradation and complex visual background interference frequently hinder detection. The study proposed OM-YOLO as an optimized model based on the YOLOv8s baseline.

To enhance detection precision and reduce the overall parameter count, the study eliminated the baseline large object detection layer and incorporated a dedicated small object detection head. Omni-Dimensional Dynamic Convolution (ODConv) was integrated into the backbone to adaptively extract rich local features from minimal pixels. In the neck architecture, a modified Receptive Field Block (mRFB) used compact dilation rates to expand the receptive area while preventing the destructive gridding effect. The architecture also integrated the Wise Intersection over Union version 3 (WIoUv3) loss function to support stable bounding box regression when processing low-quality samples through a dynamic focusing mechanism.

Empirical evaluations on the TT100K dataset showed that OM-YOLO achieved an mAP@0.5 of 90.0%, representing an improvement of 2.9% over YOLOv8s. The model achieved this accuracy while reducing the total parameter count by 58.5% to 4.6 million and maintaining a real-time inference speed of 120 FPS. Cross-dataset evaluations on the GTSDB and CCTSDB collections confirmed reliable spatial generalization capability. The study presents OM-YOLO as a viable and balanced architecture for edge device implementation in autonomous vehicles. This UDINUS activity documents institutional research addressing small traffic-sign detection, detection performance, computational efficiency, and spatial generalization in intelligent transportation systems.

Reference https://www.scopus.com/pages/publications/105042333713?origin=resultslist