Model training method, game testing method, AI character training method and device

Abstract

The present application discloses a model training method, a game testing method and a related device applied in the field of artificial intelligence. The present application includes: obtaining a game recording sample corresponding to the target game; obtaining a first predicted action corresponding to the target object by generating a network model to be trained based on the first image in the game recording sample; according to the first predicted action and game environment information , generating a first predicted image; based on the first predicted image, the first predicted action, the second image, and the second action corresponding to the second image, train the generative network model to be trained to obtain the generative network model. Based on the game strategy, the present application interacts the simulated action with the environmental information in the game, thereby generating a new game image, which facilitates traversing more game scenes, thereby improving the stability of the model.

Description

technical field [0001] The present application relates to the field of artificial intelligence, and in particular, to a model training method, a game testing method, and an AI character training method and device. Background technique [0002] Nowadays, the term artificial intelligence (AI) is mentioned more and more frequently in the field of technology, and in the field of games, AI development has become one of the most challenging tasks in game development. The effect of AI role, R & D personnel are also paying more and more attention to the quality and quantity of sample data. [0003] Currently, deep networks can be used to simulate AI characters in games based on imitation learning algorithms. First, the game recording samples are required to retain the images and corresponding actions during the game, and then input the images during the game into the deep network, and output the corresponding action labels. Based on the action labels and actual actions, the gradien...

AI Technical Summary

Problems solved by technology

[0004] However, the game images contained in the recorded game recording samples are limited, and the game scenes that can be covered are also limited. Therefore, the trained deep network lacks the ability to understand various scenes

If the AI ​​character enters a game scene that has not appeared during the recording process, it is easy to make a wrong judgment. It can be seen that the stability of the deep network is poor

Images