Real-Time Detection of Eichhornia crassipes Based on Efficient YOLOV5

The rapid propagation of Eichhornia crassipes has a threatening impact on the aquatic environment. For most small water areas with good ecology, daily manual monitoring and salvage require considerable financial material resources. Unmanned boats have important practical significance for automatic cleaning crassipes. To ensure that target can be accurately detected, we solve problems exist in lightweight model algorithm, such as low accuracy poor detection effect targets or unclear characteristics. Taking YOLOV5m 6.0 version baseline model, given computational limit real-time detection, this paper proposes to use EfficientNet-Lite0 backbone, ELU function activation function, modify pooling mode SPPF, embed SA attention mechanism, add RFB module feature fusion network improve extraction ability whole model. dataset collected hyacinth images from ponds lakes Guangxi, Yunnan, China Plant Image Library. test results show efficient YOLOV5 reached 87.6% mAP, which was 7.1% higher than YOLOV5s, average time 62 FPS. ablation experiment verifies effectiveness each YOLOV5, its parameters meet requirements unmanned boat.