Iris | Devpost

Landing page
Home page
Reality checker
Chatbot
Panic button for emergency contact alerts
Resources

Inspiration

Iris was inspired by the need to provide immediate, accessible support for individuals experiencing reality distortion or hallucinations. The goal was to create a companion app that could help users distinguish between real and perceived experiences while offering immediate support and resources.

Features

🎥 Reality Check

Uses real-time camera feed and AI to help users verify their perceptions.
Point the camera to check reality through AI-powered visual analysis.

💬 Support Chat

AI-powered conversational interface for immediate emotional support.
Real-time messaging with empathetic responses.

🏥 Medical Information

Educational resources on mental health.
Emergency contact access for crisis situations.

🚨 Emergency Alert System

Add and manage emergency contacts.
One-tap panic button to send immediate alerts.
Automated SMS alerts via Twilio.
Built using express and twilio to be pinged.

Development Process

How It Was Built

Developed using Flutter for cross-platform compatibility.
Integrated camera functionality using the camera package.
Implemented AI analysis using Nebius AI's Qwen-VL model for visual processing.
Used GetX for state management and dependency injection.
Implemented Meta’s Llama model for support chat assistance.
Designed a polished UI using Material Design with custom theming.

Challenges Faced

Switching from React Native to Flutter:
- React Native dependencies clashed, leading to a complete framework switch.
- Had to learn Dart from scratch as native development was unfamiliar.
Camera Integration:
- Handling different device orientations and camera sensors.
- Converting YUV420 format into processable images.
- Optimizing frame processing to prevent performance issues.
AI Integration:
- Managing API rate limits for real-time analysis.
- Balancing response time with accuracy.
- Handling API failures gracefully.
UX Design:
- Ensuring accessibility during distressful moments.
- Maintaining app stability and responsiveness.
- Managing permissions and device capabilities across platforms.
Model Integration:
- TensorFlow and TensorFlow Lite were outdated and incompatible.
- Finding an alternative solution took extensive research and time.
- Lost a week trying to make TensorFlow work, finishing the project in the last 5 hours.

Accomplishments

Last minute completion: finishing the project in the last 5 hours.
Real-time Visual Analysis: Successfully implemented frame-by-frame AI-based analysis.
Empathetic Chat Interface: Designed an engaging and supportive conversational experience.
Cross-Platform Compatibility: Stable functionality on both iOS and Android.
Performance Optimization: Achieved smooth camera preview and real-time processing.
User-Centric Design: Ensured accessibility during moments of distress.
First-Time Flutter Development: Completed the project within a tight deadline despite a framework switch.