Machine Learning for videogames in Unity3D
summary
Shows how to set up Unity ML-Agents 0.3 with Anaconda and TensorFlow 1.4 to train a Flappy Bird style agent from an existing Unity project, using an external brain build that exports a .bytes model then switches to the internal brain, with environment time scale sped up to 100x for accelerated training. Training on CPU takes about two to three hours to reach near‑perfect play in the demo, uses discrete actions and three scalar observations, and the exported model works when loaded back into Unity.
AI-written summary · about
chapters (5)
- 0:00Start
- 0:24What is Machine Learning?
- 1:36Machine Learning in Unity
- 14:15Setting up your machine
- 30:39Questions and Answers
related talks
36mImplementing Machine Learning the Unity way
10mMachine Learning & Unity ML-Agents update
30mBuilding, designing, teaching and training simulation environments for Machine Learning
01h 47mGamified Reinforcement Learning: Mastering AI through Video Games
48mMachine Takeover - Improving Your Game With Machine Learning & AI
32mThe Road to General AI
56mCan I do AI?
26mGames Are Smarter Than Us