This study proposes an improved tomato leaf disease detection method based on the YOLOv6 model, integrating the Convolutional Block Attention Module (CBAM) and a multi-scale feature fusion module (BiRepGFPN). The CBAM is introduced into the backbone feature extraction network to enhance the extraction of lesion features and suppress environmental interference. The BiRepGFPN is developed to improve feature fusion and enhance the localization of small lesion features. The BiRepGFPN replaces the Path Aggregation Feature Pyramid Network (PAFPN) in the YOLOv6 model to achieve effective fusion of deep semantic and shallow spatial information. Experimental results show that the model outperforms YOLOX, YOLOv5, YOLOv6, YOLOv6-s, YOLOv7, and YOLOv8 in terms of mean average precision (mAP) on the PlantDoc dataset, achieving a mAP of 93.8% on the tomato leaf disease dataset. The model demonstrates high precision (92.9%), recall (95.2%), F1 score (94.0%), and mAP (93.8%) on the tomato leaf disease dataset, showing improvements of 2.3%, 4.0%, 3.1%, and 2.7% compared to the baseline model. The study also evaluates the model's performance on the PlantDoc and FieldPlant datasets, showing significant improvements in F1 score and mAP compared to other models. The results indicate that the improved YOLOv6 model has strong detection performance and generalization capabilities in detecting tomato leaf diseases in complex environments. The study concludes that the improved model outperforms other commonly used models in terms of detection accuracy and performance. Future research directions include optimizing the model structure, designing rotated annotation boxes, diversifying the dataset, and processing multimodal data.This study proposes an improved tomato leaf disease detection method based on the YOLOv6 model, integrating the Convolutional Block Attention Module (CBAM) and a multi-scale feature fusion module (BiRepGFPN). The CBAM is introduced into the backbone feature extraction network to enhance the extraction of lesion features and suppress environmental interference. The BiRepGFPN is developed to improve feature fusion and enhance the localization of small lesion features. The BiRepGFPN replaces the Path Aggregation Feature Pyramid Network (PAFPN) in the YOLOv6 model to achieve effective fusion of deep semantic and shallow spatial information. Experimental results show that the model outperforms YOLOX, YOLOv5, YOLOv6, YOLOv6-s, YOLOv7, and YOLOv8 in terms of mean average precision (mAP) on the PlantDoc dataset, achieving a mAP of 93.8% on the tomato leaf disease dataset. The model demonstrates high precision (92.9%), recall (95.2%), F1 score (94.0%), and mAP (93.8%) on the tomato leaf disease dataset, showing improvements of 2.3%, 4.0%, 3.1%, and 2.7% compared to the baseline model. The study also evaluates the model's performance on the PlantDoc and FieldPlant datasets, showing significant improvements in F1 score and mAP compared to other models. The results indicate that the improved YOLOv6 model has strong detection performance and generalization capabilities in detecting tomato leaf diseases in complex environments. The study concludes that the improved model outperforms other commonly used models in terms of detection accuracy and performance. Future research directions include optimizing the model structure, designing rotated annotation boxes, diversifying the dataset, and processing multimodal data.