SMART INDIA HACKATHON FINALIST
ReUnite – AI System for Missing Child Identification
An AI-powered platform helping families reunite by using deep learning to match reported missing children with a centralised database.
The Overview
The Problem: In India, thousands of children go missing every day, and manual searching processes are often slow, inefficient, and resource-intensive. Police departments rely on physical descriptions and manual photo matching, which can take months or years, drastically reducing the chances of recovery.
The Solution: We built a centralised application that automates the identification process through the WCD Ministry database. The system allows citizens to scan the face of an unattended child or check the status of their filed FIR, and authorities get all the information along with the location through the app in case of any database matches. Using advanced computer vision, the platform instantly compares new images against the database of missing children to find potential matches, bridging the communication gap between the public and law enforcement. The project demonstrates the application of deep learning, optimisation and system-level thinking to a high-stakes, real-world domain.
Tech Stack
Key Features
🤖 Facial Recognition Pipeline
- Implemented CNN-based recognition using dlib and HOG descriptors to detect and recognise faces even with variations in lighting or angles.
- Optimised recognition thresholds on a highly constrained dataset (~2MB), achieving 97% match accuracy using deep metric embeddings.
📸 Real-time Inference
Enabled image capture through mobile/web cameras with live scanning and instant comparison against the WCD Ministry database.
🔔 Automated Alert and Information Routing
Deployed event-triggered notification and reporting pipelines to police, guardians, and registered NGOs based on prediction confidence.
📂 Centralised Repository
A secure dashboard that digitises records, replacing manual files and allowing cross-region data access for police.
📂 Security & System Integrity
SSL pinning, device-root detection, and mock-location blocking to eliminate spoofing and protect sensitive identities.
Why Reunite?
- Reunite works for both Android and iOS devices, enabling a large audience to use our application.
- ReUnite ensures privacy-preserving search and prevents any security and privacy breaches by not allowing citizens to browse the missing database; only model predictions are exposed.
- Geolocation tagging and Aadhaar-based authentication support traceability, minimise misuse, and accelerate actionability.
Registration Page
Challenges & Learnings
- One of the biggest technical challenges we faced was the "Single Sample" problem—where we often only had one low-quality photo of a missing child. To overcome this, we implemented image augmentation techniques (adjusting brightness, rotation, and noise) to generate robust, invariant embeddings, ensuring high recognition accuracy without needing a large training dataset.
- Researched extensively to provide a trustworthy solution and overcome major privacy and security breach challenges faced by the majority of the existing solutions, as they display all their databases to all their users.
Meet the team
Word from the mentor
