Research on the Performance of different Convolutional Neural Network Models on small Datasets

In recent image processing field, as the core of artificial intelligence, the application of machine learning methods has always occupied a very important position. The emergence of convolutional neural networks has undoubtedly played a revolutionary role. As typical models of CNN, resnet, lenet, alexnet models have all achieved great success in the work of processing large image data sets. In this paper, this paper carried out the attribute classification and recognition experiment of 800 Pokémon images using the above three complex models to explore the training effect of complex machine learning models under small data set. Besides, this analyzed the relationship between the training effect of complex models and the size of the dataset.