An object recognition method based on deep learning
Yingni Duan · 2021
Deep learning's self-learning capabilities are used to identify objects. Based on the traditional CNN convolutional neural network, this paper constructs a new network model, in which the 9 linear stacked inception modules are embedded. The use of inception modules reuses more features, improves the width of feature tensors, optimizes the training speed and improves the detection accuracy. Firstly, food picture data source is constructed and preprocessed. Then, some images are used to train the model, and the optimization algorithm SGDM is used to update the parameters, and the influence of learning rate on the training effect of the model is compared. Finally, the model is tested by testing the food pictures in the test data set, and the simulation based on MATLAB shows that the model improves the efficiency and accuracy of object recognition. It reflects the potential and advantages of deep learning in feature extraction.