DM-YOLOv8: Improved Cucumber Disease and Insect Detection Model Based on YOLOV8

Abstract

In light of the prevalent pest and disease issues faced by greenhouse cucumbers, a staple vegetable during winter, this study introduces a detection method based on the enhanced YOLOv8s model. This method aims to provide technical support for detecting and classifying pests and diseases in cucumber agricultural production. The model integrates the 'MultiCat' module for multiscale feature fusion and employs the 'C2fe' and 'ADC2f'modules to strengthen spatial and channel attention. The 'Block2d' function also facilitates the choice between average pooling and attention-based spatial pooling. Channel fusion is achieved through additive and multiplicative operations, allowing the model to delve deeper into feature learning. Experimental results confirm that our approach outperforms the original YOLOv8s model in pest detection, particularly excelling in the identification of small-scale and overlapping afflictions.

Code and data availability

The paper's cucumber disease dataset is derived from the public ai-hub 'Integrated Plant Disease Induction Data' dataset, but no authors' public URL, code deposit, or paper-specific asset is provided. The Data Availability Statement only offers contact with the corresponding author, and no allowed URL corresponds to a

No evidence-backed public reproduction asset is currently recorded.