Application of Generative Artificial Intelligence AIGC Technology Under Neural Network Algorithm in Game Character Art Design

Abstract

This paper aims to improve the efficiency and quality of game character design through artificial intelligence for generative content (AIGC). Firstly, an image extraction model based on a convolutional neural network and a game character image generation model based on a generative adversarial networks model is designed. Then, experiments are used to evaluate the loss of the model, and the performance of AIGC technology and traditional game character design methods is compared. The experimental results show that the average realism of game characters generated by the AIGC method is 0.85, which is higher than that of the traditional method of 0.82. The average value of the peak signal-to-noise ratio is 15.71, which is significantly better than the traditional method of 11.24. In addition, the Fréchet Inception Distance indicator suggests that the average of the AIGC method is 1.14, which is lower than the traditional method of 2.33. The learned perceptual image patch similarity averages 1.16, which is closer to the real sample than the traditional method of 2.17. Meanwhile, the game character design generated by AIGC technology only takes 0.85 h on average, which is much lower than the 3 h of traditional methods. Also, the uniqueness analysis of samples generated by AIGC is carried out. It is found that about 80 out of 100 generated samples are unique, indicating that the diversity of samples generated by AIGC is high, and the character design is quite different. The results show that AIGC technology has important application potential in in-game character design, which can provide higher-quality, more realistic, and diversified game character design to improve game experience and competitiveness.