Comparative analysis of YOLOv9–YOLOv12 object detection models on a custom non-open dataset
Bibliographic entry
Comparative analysis of YOLOv9–YOLOv12 object detection models on a custom non-open dataset / Q. Wang, B. Zhang, C. Wang, C. Zhang // Новые горизонты – 2025 : сборник материалов XII Белорусско-китайского молодежного инновационного форума, 27–28 ноября 2025 года / Белорусский национальный технический университет. – Минск : БНТУ, 2025. – Т. 1. – С. 102-103.
Abstract
This paper compares YOLOv9–YOLOv12 object detection models on a custom non-open dataset (6,500 images, 11 classes), all trained for 300 epochs under identical settings. Results show YOLOv12 achieves the highest mAP@0.5 (0.678), outperforming YOLOv9 (0.648), YOLOv10 (0.603), and YOLOv11 (0.649). Architectural optimizations and training stability drive performance disparities, with YOLOv12 exhibiting superior convergence and precision-recall balance, though small-object detection remains a challenge.