Ground Type Recognition Based on Lightweight Convolutional Neural Network

Aiming at the task of ground type recognition, a light-weight convolutional neural network recognition method is proposed. In order to make the lightweight convolutional model quickly fit the actual scene image, the model is fine-tuned by means of transfer learning, and the ground type image data is classified by MobileNet v2. Using the MobileNet v2 algorithm to predict on the ground type dataset, the results show that the network classification accuracy rate reaches 86%, which has a good recognition effect.