Beyond Accessibility: Engineering True Autonomy
How we are using computer vision and spatial audio to build specific tools for independence, not just aids for disability.
There is a subtle but profound difference between "accessibility" and "autonomy." Accessibility asks: "How can we make this existing system usable for a disabled person?" Autonomy asks: "How can we give this person the power to define their own interaction with the world?"
The Smart Glasses Paradigm
With our Smart Vision Glasses project, we moved away from the concept of a "medical device." We wanted to build a co-pilot. The distinction lies in the interaction model.
Traditional aids are passive or reactive (e.g., a cane hits an obstacle, you stop). Our glasses are proactive and conversational. By combining real-time object detection with a localized LLM, we allow the user to query their environment.
"Yes, about 20 meters to your left, past the blue car."
Latency as a Safety Metric
For a blind user navigating a busy Lagos street, latency is not just an inconvenience; it's a safety hazard. Sending video data to the cloud, processing it, and sending audio back takes too long.
This is why on-device inference is non-negotiable. We optimized our vision models to run heavily quantified on mobile-grade processors. We achieved a glass-to-ear latency of under 200ms. This speed allows the system to warn of a sudden obstacle—a motorcycle, an opening door—in real-time.
We are not just building gadgets. We are restoring the agency that the physical world sometimes takes away.