The invention provides an end-to-end network interface detection method based on deep learning. The method comprises the following steps: (1) obtaining switch network interface images of different backgrounds and different cabinets of a machine room; (2) classifying the network interface images of the switch; Marking each image, and making a data label; (3) manufacturing a switch network interfaceimage data set by using a data enhancement technology; And (4) training a deep network YOLO v3 model, and storing a model result. The internet access image is obtained from the camera of the intelligent indoor inspection robot, the position and the category of the internet access in the image are obtained through algorithm processing, and then on-site real-time deployment is achieved. According to the invention, the network interface detection accuracy of the switch in the machine room can be improved, so that the deployment efficiency of the intelligent inspection robot is improved.