This paper addresses the challenge of detecting tomato leaf diseases in natural environments, where factors such as lighting changes, occlusions, and small lesion sizes pose significant detection challenges. The study proposes an improved tomato leaf disease detection method based on the YOLOv6 model, incorporating attention mechanisms and multi-scale feature fusion. The key contributions are: 1. **Convolutional Block Attention Module (CBAM)**: Integrated into the backbone feature extraction network to enhance lesion feature extraction and suppress environmental interference. 2. **Multi-Scale Feature Fusion Module (BiRepGFPN)**: Based on the Reparameterized Generalized Feature Pyramid Network (RepGFPN), it captures the characteristics of small lesions and improves feature fusion expression. 3. **Model Evaluation**: The improved model outperforms the baseline YOLOv6 model and other mainstream models (YOLOX, YOLOv5, YOLOv7, YOLOv8) on the PlantDoc dataset, achieving mean average precision (mAP) improvements of 7.7%, 11.8%, 3.4%, 5.7%, 4.3%, and 2.6%, respectively. 4. **Performance on Tomato Leaf Disease Dataset**: The model demonstrates a precision of 92.9%, a recall rate of 95.2%, an F1 score of 94.0%, and an mAP of 93.8%, showing improvements of 2.3%, 4.0%, 3.1%, and 2.7% compared to the baseline model. The proposed method effectively enhances the detection accuracy and generalization capabilities of tomato leaf disease detection, making it suitable for real-world applications in agriculture.This paper addresses the challenge of detecting tomato leaf diseases in natural environments, where factors such as lighting changes, occlusions, and small lesion sizes pose significant detection challenges. The study proposes an improved tomato leaf disease detection method based on the YOLOv6 model, incorporating attention mechanisms and multi-scale feature fusion. The key contributions are: 1. **Convolutional Block Attention Module (CBAM)**: Integrated into the backbone feature extraction network to enhance lesion feature extraction and suppress environmental interference. 2. **Multi-Scale Feature Fusion Module (BiRepGFPN)**: Based on the Reparameterized Generalized Feature Pyramid Network (RepGFPN), it captures the characteristics of small lesions and improves feature fusion expression. 3. **Model Evaluation**: The improved model outperforms the baseline YOLOv6 model and other mainstream models (YOLOX, YOLOv5, YOLOv7, YOLOv8) on the PlantDoc dataset, achieving mean average precision (mAP) improvements of 7.7%, 11.8%, 3.4%, 5.7%, 4.3%, and 2.6%, respectively. 4. **Performance on Tomato Leaf Disease Dataset**: The model demonstrates a precision of 92.9%, a recall rate of 95.2%, an F1 score of 94.0%, and an mAP of 93.8%, showing improvements of 2.3%, 4.0%, 3.1%, and 2.7% compared to the baseline model. The proposed method effectively enhances the detection accuracy and generalization capabilities of tomato leaf disease detection, making it suitable for real-world applications in agriculture.