MSNav: Zero-Shot Vision-and-Language Navigation
MSNav is a breakthrough framework that addresses the limitations of traditional intelligent agents in zero-shot vision-and-language navigation. The system features dynamic topological memory and spatial capabilities, enabling agents to actively filter relevant information for tasks rather than passively receiving all information.
Key Features:
- Dynamic topological memory system
- Spatial reasoning capabilities
- Modular framework design
- Zero-shot navigation performance
Technologies: Python, PyTorch, Computer Vision, Natural Language Processing
Status: Under Review, Preprint Available