IMAGE IDENTIFICATION METHOD AND SYSTEM BASED ON CONVOLUTIONAL NEURAL NETWORK, AND ELECTRONIC DEVICE

摘要

Disclosed are an image identification method and system based on a convolutional neural network, and an electronic device. The method comprises: inputting data of an image to at least two serially-connected convolutional layers for feature extraction to obtain extracted feature data, wherein the kernel sizes of the convolutional layers are not greater than 5×5; performing feature data dimensionality reduction and extraction on the extracted feature data by means of pooling layers and the convolutional layers, to obtain dimension-reduced feature data, wherein the pooling layers use average pooling; inputting the dimension-reduced feature data of the image to a fully-connected layer to obtain two-dimensional feature values of the data of the image; classifying the two-dimensional feature values by means of a classifier to obtain an identification result of the image. Also disclosed is an image identification system based on a convolutional neural network. The image identification method and system based on a convolutional neural network extract feature data by means of convolutional layers having smaller kernels, so as to better and quickly extract local features of an image, thereby improving the speed and efficiency of image identification.