TOM-YOLO: Tomato Leaf Disease Detection Using Enhanced YOLOv12 Framework
Students & Supervisors
Student Authors
Supervisors
Abstract
Tomato leaf diseases significantly impact agricultural productivity, making early and accurate detection essential for sustainable crop management. In this work, we propose TOM-YOLO, a robust and high-accuracy framework built upon YOLOv12, designed to address the challenges of tomato leaf disease detection under complex field conditions. The proposed TOM-YOLO enhances YOLOv12 for tomato leaf disease detection by replacing standard convolutions with GhostConv layers at positions 3, 15, and 19 and incorporating A2C2f blocks to improve feature representation and detection accuracy. An SPPF module at layer 9 captures multi-scale context for improved detection of small and large lesions, while C2f modules at layers 2, 6, and 21 in the backbone and neck extract fine-grained lesion patterns with attention. Finally, a custom detection head is designed to classify nine tomato leaf diseases into nine categories. Experimental results demonstrate that TOM-YOLO achieves superior performance compared to the original YOLOv12, with precision, recall, F1-score, mAP@50, and mAP@50-95 reaching 87.5%, 81.1%, 84.0%, 90.4%, and 77.9%, respectively.
Keywords
Publication Details
- Type of Publication:
- Conference Name: IEEE International Conference on Robotics, Automation, Artificial-Intelligence and Internet-of-Things 2025
- Date of Conference: 27/11/2025 - 27/11/2025
- Venue: Military Institute of Science and Technology, Dhaka, Bangladesh
- Organizer: IEEE Bangladesh Section and IEEE Robotics and Automation Society, Bangladesh chapter