s

street-fighter-ai

AI agent for beating Street Fighter II using deep reinforcement learning.

AI AgentOpen SourceGrowing
Visit Website6.5K

What is street-fighter-ai?

street-fighter-ai is aI agent for beating Street Fighter II using deep reinforcement learning.

About

Street Fighter AI is an AI agent designed to defeat the final boss in 'Street Fighter II: Special Champion Edition' using deep reinforcement learning. It operates by analyzing the game's RGB pixel values and has achieved a 100% win rate in the first round of the final level. This tool is ideal for developers interested in game AI and reinforcement learning techniques.

Strengths

  • Achieves high win rates against game bosses.
  • Utilizes popular libraries for reinforcement learning.
  • Includes detailed training logs and performance metrics.
  • Supports model training and testing with ease.
  • Open-source and community-driven project.

Limitations

  • Requires legal acquisition of game ROM files.
  • Setup can be complex for beginners.
  • Performance may vary significantly based on model training.
  • Limited to a specific game and scenario.
  • Overfitting issues may affect generalization.

Use Cases

Train AI agents to play retro video games.Analyze AI performance using Tensorboard.Experiment with different reinforcement learning models.Capture and record gameplay for analysis.Modify and enhance AI strategies for improved performance.

Integrations

OpenAI Gym RetroStable-Baselines3TensorboardAnacondaffmpeg