Discover the fascinating world of artificial intelligence in game development
- Module type: Elective module
- Duration period: Summer semester, winter semester
- Type of examination: Project work with three interim presentations and a final presentation
- Demands: 4th semester
Module description
This course provides a comprehensive introduction to the fascinating world of Artificial Intelligence (AI) in game development. Students will learn about the possible applications of AI and immerse themselves in areas such as deep learning, reinforcement learning, generative models, prompt engineering as well as augmented and virtual reality. Practical units include the implementation of neural networks, image and text classification, training of AI agents, use of generative models and the use of various frameworks and tools. The project topic includes the development of a game with NPC training, the use of AI models for dialogs and image generation as well as the development of games for augmented reality.
Lecture units
- Application prospects of AI in game development
- Introduction to Deep Learning
- Introduction to reinforcement learning (RL)
- Introduction to generative models and prompt engineering
- Introduction to Augmented and Virtual Reality
Practical Training units
Implementation of simple neural networks
- Classification of digits on images (MNIST dataset)
- Photo classification (CIFAR-10 dataset)
Training of an RL-based agent
- Robot Vacuum Cleaner example
- Lunar Lander Console game
Use of generative models (Language Models, Text2Image Models)
- Prompt Engineering with ChatGPT and Stable Diffusion
Use of various frameworks / tools
- Git and Gitlab, Docker, Kubernetes, Weights & Biases
Project topics
Development of a game and training of NPCs
- Gym(nasium) is a framework for training agents using reinforcement learning
- Existing environments can be used (e.g. Atari Games)
- Python; Keras (Tensorflow) / Pytorch; high school, Numpy
Development of a text adventure
- Use of language models for dialogs
- Use of CV models for image generation
- Python; Transformers / OpenCV; OpenAI / Midjourney; FastAPI
Unity Machine Learning Agents
- ML Agents is a framework for training agents in Unity
- NPCs can be trained and play individually, against each other or against humans
- C#, Python, Unity
Games, AI and Augmented Reality
- Development of a game in Unity (Exit Game, Geocaching, Penalty Shootout)
- Deployment on HoloLens2 or smartphone
- C#, Python, Java; Unity or Android Studio; Mixed Reality ToolKit or ARCore
