Tiresias

• 

  Tiresias

Inspiration

In Greek mythology, Tiresias was a blind prophet who could see what others couldn't. This inspired us to build technology that gives visually impaired individuals a new way to "see" their surroundings.

Over 2.2 billion people globally have vision impairment. While guide dogs and white canes remain essential tools, they have limitations — they can't warn you about overhead obstacles, identify objects, or describe your environment. We asked: What if your iPhone could become an intelligent guide that perceives the world for you?

The convergence of powerful on-device AI, LiDAR sensors in modern iPhones, and real-time edge computing made this vision possible.

What it does

Tiresias transforms an iPhone into an intelligent navigation assistant:

• Real-time obstacle detection — Uses the camera and LiDAR to identify objects, people, and hazards in your path
• Depth sensing — LiDAR measures precise distances to obstacles (walls, furniture, stairs, curbs)
• Voice guidance — Natural language descriptions: "Person approaching from the left, 3 meters" or "Door on your right"
• Haptic feedback — Distinct vibration patterns warn of obstacles:
  • Intensity increases as objects get closer
  • Different patterns for different danger levels
  • Directional cues (left/right vibrations)
• Edge processing — Streams video to a local edge server for AI inference, minimizing latency
• Privacy-first — All processing happens locally, no cloud uploads

How we built it

iOS App (Swift/SwiftUI):

• Camera and LiDAR capture pipeline using AVFoundation and ARKit
• WebSocket streaming for real-time video transmission
• Haptic engine integration using Core Haptics for nuanced feedback
• VoiceOver-compatible UI with accessibility-first design
• Network resilience with WiFi/USB fallback connections

Edge Server (Python/FastAPI):

• Real-time object detection using computer vision models
• Depth map processing from LiDAR data
• Spatial reasoning to determine obstacle positions and trajectories
• Natural language generation for voice guidance
• WebSocket server for low-latency bidirectional communication

AI/ML Pipeline:

• Object detection for identifying obstacles, people, vehicles
• Depth estimation fusion (LiDAR + monocular depth)
• Path analysis to determine safest walking route
• Priority system to warn about most critical obstacles first

Track Submission:

• Social Good
• Machine Learning / AI

Built With

• ai
• fastapi
• machine-learning
• python
• react
• swift
• websockets
• yolo