DogeAI is an AI model of a dog that learns and adapts in real-time through visual data. Every N steps, a reward model is triggered to detect if the virtual dog has hit an obstacle, making it a self-learning technology. Developed entirely in C++ from scratch, open-source and always available for non-commercial use.
Demo features a virtual dog running in Unreal Engine enviroment
- Optimization: We're constantly working to improve the performance of DogeAI
- Cross-platform build: We aim to make DogeAI available across multiple platforms, including STM32
- Efficient weights export