Vegetable Disease and Pest Target Detection Algorithm Based on Improved YOLO v7

In response to the common issues of vegetation cover and similar disease features in vegetable plant detection in real environments, which can lead to missed and false detections, this paper proposes an improved cabbage pest detection algorithm based on YOLOv7. Firstly, a data augmentation module is used to enhance the dataset, simulating data features in natural environments such as occlusion, low brightness, flipping, and translation. Secondly, the AFP module is used as the neck network to enhance the model's adaptive feature fusion ability in the feature fusion area; Introducing the EMA attention mechanism module to further aggregate pixel level features through dimensional interaction, reducing missed and false detections of downy mildew and leaf miner in natural vegetation environments. The experimental results show that compared to YOLOv7, the Improved YOLOv7 pest and disease detection algorithm has an accuracy of 94.2% on the AI studio dataset, an improvement of 25.9% compared to the original algorithm, and an improvement of 4.6% in FPS compared to the original algorithm, achieving improvements in detection accuracy and speed. The pest and disease detection is faster and more accurate