Identification of shanlan upland rice varieties based on improved YOLOv8s

As a unique wild upland rice in Hainan, shanlan upland rice has important germplasm resources value. In order to deal with the problem of shanlan upland rice variety recognition, this paper proposes a shanlan upland rice variety recognition algorithm based on improved YOLOv8 s. Firstly, the attention mechanism and CBAM ( Convolutional Block Attention Module ) module are introduced into YOLOv8s to enhance the ability of the model to extract key features in the task of identifying Shanlan rice varieties, so as to improve the accuracy and robustness of the model. Secondly, by adding SE ( Squeeze-and-Excitation ) module to YOLOv8 s, the utilization efficiency of the model to features is effectively improved. The experimental results show that: in terms of accuracy, the YOLOv8s-attention-CBAM model with attention mechanism and CBAM module is 89.8 %, which is 1 % higher than the YOLOv8s-Attention-SE model with SE module and 5.44 % higher than the original YOLOv8s model; the YOLOv8s-Attention-SE model shows more stable performance, and its average accuracy fluctuates less. The improved YOLOv8s model can be better applied to the identification of shanlan upland rice varieties.