Objective An algorithm was developed based on YOLOv8s to accurately identify the pests like Trichophysetis cretacea and thrips on jasmine plants.

Method A P2 layer for detecting small objects and the mechanism of coordinate attention (CA) and multi-scale feature fusion (MSFF) for integrating information were incorporated into the original YOLOv8s model to improve the detection and identification of the pest insects, such as T. cretacea and thrips, that commonly infested jasmine plants. Furthermore, a web management platform and a WeChat mini-program were added to the program for easily accessible visual display and remote viewing.

Result The modified algorithm raised mAP_0.5 by 2.94% over the original model, showing an 1.97% increased rate of corrective recognition on T. cretacea and 4.38% on thrips, as well as a 0.19% raised recall rate on T. cretacea and 0.35% on thrips.

Conclusion By embracing the size factor and additional computation loads, the modified algorithm improved the accuracy in detecting and recalling T. cretacea and thrips under complex conditions. The online accessibility provided the program with convenient data visualization and remote monitoring for the pest control at jasmine gardens.