M-YOLO v8s: Classification and Identification of Different Microalgae Species Based on the Improved YOLO v8s Model for Prevention of Harmful Algal Blooms
Jianhong Dong, Junsheng Wang, Huimei Lin, Wen Liu
Abstract
To achieve rapid and accurate classification and identification of different microalgae species, we developed the M-YOLO v8s model based on the YOLO v8s model, replacing the C2F module with the C2F_Faster module in the backbone to achieve a lighter network structure and efficient feature extraction, and Focal-SIoU loss was introduced to enhance the stability of the model. SRGAN was employed to process the microalgae images captured by a microscope to increase the diversity of the data set before training to improve the robustness of the model. The detection accuracy and speed of M-YOLO version 8 were significantly improved, while the complexity was reduced. The precision increased from 98.5 to 98.9%, and the recall was 99.1%. Furthermore, Params decreased from 11.13 to 8.31 million and FLOPs decreased from 28.4 to 21.4 billion, indicating that fewer computing resources are required. The improved M-YOLO v8s model is crucial for the early warning and prevention of harmful algal blooms.