TrackWay
An offline assistive system combining local language models, computer vision, and voice interaction on a Raspberry Pi.
Status: Assistive system / prototype
Engineering problem
Bring assistive interaction to constrained hardware without making the offline implementation depend on a cloud service.
Approach & architecture
- Quantized language models, YOLO, and voice interaction deployed locally on Raspberry Pi and Linux.
- Rust/C++ and Python across the device-side implementation.
- A separate hybrid prototype explored microcontrollers, ultrasonic sensors, and Gemini function calling. That cloud-dependent experiment is distinct from the offline system.
Outcomes & scope
- Offline assistive implementation developed for Raspberry Pi.
- The separate cloud-assisted prototype is documented as a different implementation.