Scene 1 (0s)
[Virtual Presenter] Arm AI Developer Challenge NeuroLens Hybrid Intelligence for the Visually Impaired.
Scene 2 (13s)
[Audio] The Problem & The Vision The "Speed vs. Smarts" Dilemma The Gap: Blind users currently choose between fast, dumb apps (safe but vague) or smart, slow apps (detailed but laggy). The Danger: A 3-second delay in identifying a moving car or obstacle is a safety hazard. Our Vision: "Bionic Vision" — combining the instant reflexes of a biological eye with the understanding of a human brain. The Goal: Zero-latency safety + high-fidelity understanding in one app..
Scene 3 (47s)
[Audio] The Solution - Hybrid Intelligence NeuroLens: The Best of Both Worlds Reflex Mode (Local): Runs 100% on-device (Arm CPU/GPU) for <100ms latency and instant obstacle detection. Cognitive Mode (Cloud): Leverages Mistral AI for rich, poetic scene descriptions when the user asks for detail. Seamless Handover: The local layer handles safety; the cloud layer handles understanding. True Offline: Core safety features work perfectly in subways, elevators, and remote areas..
Scene 4 (1m 9s)
[Audio] Under the Hood - Built on Arm Optimized for Mobile Edge Arm-First Design: Leveraged `tfjs-react-native` to accelerate the SSDLite model via the Arm GPU (WebGL). Smart Bundling: We bypassed standard asset limits by bundling model weights directly into the binary for instant startup. Privacy by Design: Video frames never leave the device unless the user explicitly requests a cloud description. Efficiency: Running a lightweight MobileNet V2 model ensures all-day battery life..
Scene 5 (1m 49s)
[Audio] Demo & Impact Empowering Independence Real-Time Safety: Haptic feedback vibrates instantly when obstacles are detected, preventing accidents. Voice Interface: Users can converse with their environment ("What's in front of me?") using natural language. Accessibility: Designed with high-contrast UI and screen-reader compatibility from day one. Immediate Impact: Transforms a standard smartphone into a powerful, intelligent mobility aid..
Scene 6 (2m 21s)
[Audio] The Future of NeuroLens Roadmap & Next Steps Personalized AI: Teach NeuroLens to recognize specific items like "my keys" or "my dog." Spatial Audio: Use 3D sound to place objects in the user's ear (e.g., hear a beep on the left if the door is on the left). Indoor Navigation: Guide users to bathrooms or exits in unmapped buildings using visual landmarks. Wearable Port: Porting the "Reflex Layer" to smart glasses powered by low-power Arm Cortex-M chips..