External damage hidden danger source image recognition method and device, medium and terminal equipment

The invention discloses an external damage hidden danger source image recognition method and device, a medium and terminal equipment. The method comprises the following steps: introducing an ECA attention mechanism to improve weights corresponding to different channel features in a YOLO network; training the YOLO network by using the marked external damage hidden danger source image data set to obtain a target parameter and a target weight of the YOLO network; performing form conversion on the trained YOLO network by using a TensorRT optimizer; and deploying the converted YOLO network to an identification terminal, and identifying an external damage hidden danger source image through the identification terminal. According to the invention, the YOLO network after structure improvement, data training and model conversion is deployed to the identification terminal to detect the external damage hidden danger source in real time, and the problem of current external damage hidden danger source identification is eff