Affordability
Keep the full system below premium assistive-device pricing.
Commodity compute, repurposed cells and modular components reduce cost without forcing cloud dependence.
Field controls
Numerically integrated ray-bending approximation with a procedural relativistic accretion disk. Visual model—not a full Kerr-metric solver.
CASE STUDY / 01
Vision translated into action.
JØ7 Vireo is a low-cost, offline assistive-vision platform for blind and visually impaired users. It combines object detection, distance awareness, speech and haptic feedback on edge hardware.
[01] SYSTEM ARCHITECTURE
Sense. Interpret. Communicate.
The system is designed so perception never ends as a bounding box on a screen. It must become useful guidance.
01Capture
Camera and distance sensors observe the near environment.
02Infer
Object detection runs locally on Raspberry Pi and Hailo acceleration.
03Prioritise
Software decides what is relevant enough to interrupt the user.
04Respond
Speech and a haptic belt communicate direction and urgency.
86,000Images in the author-confirmed training corpus.
26TOPS available through the Hailo-8L accelerator.
OfflineNo cloud dependency for core perception and feedback.
Privacy
Video should not need to leave the device.
Local inference protects environmental imagery and keeps the system functional when connectivity disappears.
Cognitive load
More detections can make the product worse.
Useful feedback requires prioritisation, timing and restraint rather than narrating every object in view.
Failure
Safety signals need redundancy.
Visual inference, distance sensing and haptics cover different weaknesses instead of pretending one sensor is infallible.