Implementation of Artificial Intelligence Anime Stylization System Based on PyTorch

In the field of computer image vision, with the rise of artificial intelligence and deep learning, the image style migration techniques used are becoming more and more sophisticated. To address the problem that style migration based on traditional techniques has strong limitations and one procedure can only implement a single scene, this paper first discusses the model structures of ANN (artificial neural network) and MLP (multi-layer perceptron), and the loss functions, and then the design and implementation of an artificial intelligence animation stylized system based on ANN and MLP is proposed. Through the test results of the system, it can be verified that the algorithm implementation used in the system corresponds to multiple types of image styles, and the results of migration conversion can also achieve a high degree of agreement with the original image samples, finally A system for converting multiple image styles has been implemented.