🏠 Shelter Safety AI

A Django web application integrating multiple AI models to ensure a safe and stable daily life for children in shelters.

Developed as part of the 3IA Program — Challenge-Based Learning Project at Esprit School of Engineering.

---

Overview

Big Idea: Child Safety
Essential Question: How can we ensure that children stay safe every day in shelters?
Challenge: Provide a safe and stable daily life for children in shelters.

This platform combines 7 AI objectives deployed through a unified Django backend, providing real-time risk detection and alerting for shelter staff.

---

Features

• Sanitation risk detection — Identifies cockroaches, trash, and wet surfaces using YOLO11s (mAP50: 0.819)
• Structural damage assessment — Detects spalling and efflorescence
• Overcrowding prediction — Estimates occupancy and generates density heatmaps
• Accident hazard detection — Identifies dangerous objects and missing safety equipment
• Behavioral risk detection — Classifies risky child behaviors and computes risk scores
• Allergen detection — Identifies allergenic substances in food ingredients
• Meal nutrition analysis — Analyzes meals for nutritional balance

---

Tech Stack

Backend

• Python 3.12
• Django
• Django REST Framework

AI Models

• PyTorch 2.10
• Ultralytics YOLO11s
• OpenCV

Explainable AI

• LIME
• Occlusion Saliency Maps

Deployment

• Heroku (via GitHub Education)

---

Directory Structure

Getting Started

git clone https://github.com/your-username/shelter-safety-ai
cd shelter-safety-ai
pip install -r requirements.txt
python manage.py migrate
python manage.py runserver

Acknowledgments

Developed under the guidance of professors at Esprit School of Engineering.
CBL Project — 3IA Program 2025/2026.