AI agents for the boardgame Splendor

Bot Pokémon Gen 1 (Random Battle Showdown) entraîné en RL avec MaskablePPO, imitation + curriculum + self-play ; suivi MLflow.

Autonomous Clash Royale agent using computer vision and reinforcement learning. Dual YOLOv11 models + SORT tracking perceive the game state from screen pixels; a Maskable PPO policy decides card placement in real time via simulated mouse clicks.

RL agent learns to furnish a small bedroom. MaskablePPO on randomized rooms with a multiplicative reward.

Unofficial Stable-Baselines3 extension package with MaskableRecurrentPPO.