摘要：In recent years there have been many successes of using deep learning for imaging classification recognition. In this work firstly we discuss in details the differences between machine learning and deep learning from the limitations of traditional machine learning, and gives a detail introduction to the advantages of typical deep convolution neural network in image classification. Deeper neural networks are more difficult to train, this paper presents an improving deep learning convolutional neural network (CNN) structure model and gain accuracy from considerably increased depth. We also show that this improving structure model leads to the prediction results are higher than the original deep-learning CNN structure model with training and testing on the published data set.
