Tomato Leaf Disease Detection Method Based on Multi-Scale Feature Fusion.

Abstract

Tomato is a key economic crop whose yield and quality depend heavily on the early and accurate detection of leaf diseases. Conventional diagnosis based on manual observation is labor-intensive and prone to subjective bias. To overcome the limitations of disease detection under complex environmental conditions, this study presents an enhanced YOLO11n-based detection framework for tomato leaf diseases. The proposed model integrates an EfficientMSF module in the backbone to strengthen multi-scale feature extraction, introduces a C2CU module to enhance global contextual representation, and employs a CAFMFusion module to achieve efficient fusion of local and global features. Experiments were conducted on a self-constructed dataset containing nine tomato leaf categories, including eight disease types and healthy samples. The proposed approach achieves an average Recall of 71.0%, mAP@0.5 of 76.5%, and mAP@0.5-0.95 of 60.5%, outperforming the baseline YOLO11n by 3.4%, 1.3%, and 2.0%, respectively. In particular, for the challenging Leaf Mold class, mAP@0.5 improved by 3.4%. These results demonstrate that the proposed method possesses strong robustness and practical applicability in complex field conditions, offering an effective solution for intelligent tomato disease monitoring and precision agricultural management.

Plant phenotyping relevance

トマト葉の病徴・病害状態を画像から検出する深層学習法を開発し、データセット上で性能比較・検証しているため、植物フェノタイピング手法が中心である。

Code and data availability

The paper uses a self-constructed dataset of 2212 tomato leaf disease images (Sanya, Hainan) with LabelImg annotations, but no public deposit or URL is provided. The Data Availability Statement says the dataset is available from the corresponding author upon reasonable request. No author analysis code, trained model,或

No evidence-backed public reproduction asset is currently recorded.