SeeThrough:

Inspiration

We started with a simple question:
Why should blind individuals be unable to perceive their surroundings when technology can bridge the gap?

Current assistive technologies either lack real-time responsiveness, require bulky hardware, or fail to provide intuitive feedback. With AI-powered object detection, text reading, and proximity alerts, we set out to build a lightweight, real-time solution that enhances independence and safety.

What It Does

SeeThrough detects objects in front of visually impaired users and describes them aloud using text-to-speech. It can:

Identify objects (e.g., "Chair on your left.")
Read printed text on command
Warn about obstacles using proximity sensors

How We Built It

YOLO for real-time object detection
ESP32-CAM for live video streaming
Ultrasonic Sensors for proximity detection
Text-to-Speech (TTS) for spoken feedback
Edge AI Optimization for low-latency performance

Challenges We Ran Into

Reducing latency for real-time object detection
Processing AI models on low-power hardware
Ensuring clear speech feedback without overwhelming users

Accomplishments We're Proud Of

Successfully implemented real-time object detection
Optimized text reading accuracy for better accessibility
Developed a compact and user-friendly prototype

What We Learned

Optimizing AI for embedded devices is crucial for speed and efficiency.
User-friendly feedback matters—balancing information with clarity.
Hardware limitations require smart optimizations for performance.

What’s Next for SeeThrough?

Integrating Ray-Ban smart glasses or VR headsets for a hands-free experience.
Optimizing AI models for greater detection accuracy and efficiency.