Revolutionizing tomato disease detection in complex environments.

Area of Science:

• Agricultural technology
• Computer vision
• Plant pathology

Background:

• Tomato leaf diseases pose a significant challenge to manual detection in agriculture.
• Existing automated detection algorithms often struggle to balance speed and accuracy, particularly for small-scale diseases in varied conditions.
• There is a need for efficient and precise automated systems for early disease identification in crops.

Purpose of the Study:

• To develop an innovative, high-precision, and lightweight detection algorithm for tomato leaf diseases.
• To address the limitations of current algorithms in terms of speed, accuracy, and computational complexity.
• To improve the early detection and management of plant diseases in agricultural settings.

Main Methods:

• Introduced FCHF-DETR (Faster-Cascaded-attention-High-feature-fusion-Focaler Detection-Transformer), a novel algorithm based on RT-DETR-R18.
• Utilized FasterNet as the backbone, Cascaded Group Attention, High-Level Screening-feature Fusion Pyramid Networks (HSFPN), and Focaler-CIoU loss function.
• Trained and evaluated the model on a dataset of 3147 RGB images of tomato leaf diseases.

Main Results:

• FCHF-DETR achieved high performance metrics: 96.4% Precision, 96.7% Recall, 89.1% mAP50-95, and 97.2% mAP50.
• The algorithm demonstrated significant reductions in computational complexity: 9.2G FLOPs and 3.6M parameters.
• The proposed modifications enhanced detection accuracy without substantial impact on efficiency.

Conclusions:

• The FCHF-DETR algorithm effectively improves the accuracy and efficiency of tomato leaf disease detection.
• The study successfully addressed the dual challenges of precision and computational complexity in automated plant disease identification.
• This lightweight and high-performance model offers a promising solution for real-world agricultural applications.