InfoNav Demo Video
This release includes a demonstration video showing InfoNav's capabilities in semantic-aware object navigation.
Features Demonstrated
- Vision-Language Model (VLM) based object detection
- LLM-based semantic reasoning
- Unified value framework combining semantic value and information gain
- Zero-shot navigation in indoor environments
Video
See the attached InfoNav.mp4 for a complete demo of the system in action.
System Overview
InfoNav is an intelligent navigation system for embodied agents in indoor environments, integrating:
- GroundingDINO and BLIP2 for open-vocabulary detection
- DeepSeek LLM for semantic reasoning
- Frontier-based exploration with semantic value maps
- A* path planning with TSP optimization
For installation instructions and usage details, please see the README.