Project Dristi

iSight: A smart cane delivering precise obstacle detection and intuitive alerts, seamlessly paired with 'Dristi'—the intelligent app guiding visually impaired users safely.

Comment

Inspiration

I am from Jhapa, Nepal, and I grew up with a neighbor who was completely blind yet remarkably tech-savvy. He managed his own YouTube channel and was actively trading cryptocurrencies as early as 2018. Despite his technological skills, he constantly expressed frustrations about the poor accessibility of roads and the unavailability of affordable, user-friendly tech designed specifically for visually impaired individuals. Witnessing his struggles firsthand inspired us deeply, motivating us to create a technological solution. Additionally, commercially available smart canes typically cost over $1000, prompting us to build an affordable alternative under $100.

What it does

iSight is an intelligent smart cane equipped with ultrasonic sensors, vibration motors, and audio feedback to detect obstacles and alert users in real-time. Integrated with Dristi, our companion web app, it provides additional functionalities such as voice-guided navigation, location sharing, hazard alerts, SOS calls, and conversational AI support through voice interactions. Both the app and the cane work independently, similar to how an Apple Watch and an iPhone function, each useful alone but more powerful together.

How we built it

Hardware:

  • Raspberry Pi 4 (model B) as the processing core.
  • Plain white cane.
  • Ultrasonic sensors (HC-SR04) for obstacle detection.
  • Haptic motors and buzzer for tactile and auditory feedback.
  • GPS module (Neo 6M) for live location tracking.

Software:

  • Frontend built with React, integrating Google Maps SDK for navigation and visualization.
  • Backend powered by Flask and Firebase for efficient data handling and real-time communication.
  • AI components leveraging Google Gemini for conversational intelligence.

Challenges we ran into

  • Communication and synchronization between the smart cane hardware and the Dristi app.
  • Whisper AI model was too large, making deployment difficult within the constraints of a free hosting plan and limited timeframe.
  • Achieving reliable real-time obstacle detection with minimal latency.
  • Balancing accuracy and responsiveness in voice-controlled features.
  • Initially planning to use an 18650 Li-ion battery (5000 mAh), we discovered it provided only 3.7V, insufficient for the 5V required by Raspberry Pi.
  • The lidar we initially bought had 12 ports, but the Raspberry Pi only had 8 ports, leading us to opt for Ultrasonic sensors (HC-SR04) instead.

Accomplishments that we're proud of

  • Successfully built a fully functional hardware prototype within tight time constraints.
  • Integrated multiple technologies seamlessly to provide intuitive user experiences.
  • Created a high-quality, user-friendly frontend that visually represents real-time cane feedback.
  • Achieved ultra-low cost for the smart cane, reducing expenses by more than 90% compared to existing commercial products.

What we learned

  • As beginners, we significantly enhanced our understanding and practical skills in hardware development.
  • Effective collaboration between hardware and software teams requires continuous communication.
  • Real-time applications demand careful consideration of performance optimization.
  • The importance of user-centric design, especially when building assistive technologies.

What's next for Project Dristi

  • Making the smart cane lighter and more practical for everyday use.
  • Expanding the web app functionality to include more advanced AI-driven environmental awareness and additional helpful features.
  • Field-testing with visually impaired users to gain feedback and improve usability.
  • Exploring additional sensor integrations (like lidar or cameras) to further increase safety and accuracy.

Built With

Share this project:

Updates