ADHARA

AI-Powered Learning Friction Detection System

Precision AI for early detection of learning difficulties through multimodal behavioral analysis

---

🎯 Overview

ADHARA (AI-Driven Holistic Assessment for Developmental Recognition and Assistance) is an intelligent early warning system that detects learning friction patterns in children. By analyzing behavioral signals across multiple modalities, ADHARA helps educators identify potential learning difficulties like Dyslexia, Dyscalculia, and ADHD before they become persistent challenges.

💡 Why ADHARA? Traditional assessments only measure results. ADHARA analyzes the process of learning—capturing hesitation, stress, attention patterns, and cognitive load in real-time.

✨ Key Features

• 🧠 Adaptive Learning Sessions - Questions adjust in real-time based on performance patterns
• 👁️ Face & Emotion Analysis - Detects stress, confusion, and engagement through webcam
• 🖱️ Mouse Dynamics Tracking - Analyzes hesitation, jitter, and interaction velocity
• 🎤 Speech Analysis - Monitors fluency, stammering, and verbal hesitation patterns
• 📊 AI-Powered Reports - Generates clinical-grade analysis using local LLM (Ollama)
• 🔒 Privacy-First - All processing happens locally, no data leaves the machine

---

📈 Multimodal Analysis

ADHARA captures four distinct behavioral streams to create a comprehensive friction profile:

Data Stream	What We Measure	Technology
Vision	Gaze tracking, blink rate, attention drift	face-api.js
Emotion	Micro-expressions, stress detection, engagement	TensorFlow.js
Motor	Mouse velocity, hesitation patterns, jitter	Custom tracking
Voice	Fluency, filler words, stammering, pauses	Web Speech API

---

🛠️ Installation & Setup Guide

Prerequisites

Before you begin, ensure you have the following installed:

1. Node.js (v18 or higher) - Download Here
2. Git - Download Here

1. Close the Repository

git clone https://github.com/JithuMon10/ADHARA-AI-Powered-Learning-Friction.git
cd ADHARA-AI-Powered-Learning-Friction

2. Install Frontend Dependencies

Important: You must move into the client directory before installing dependencies.

cd client
npm install

3. Start the Application

npm run dev

The application will launch at http://localhost:5173.

---

🤖 Setting Up AI Analysis (Ollama)

For the detailed "Clinical Analysis Report" to work generates, you need Ollama running locally. The app works without it, but the "Analysis" tab will be disabled.

Recommended Model: Qwen 2.5 7B

We recommend using Qwen 2.5 7B for the best balance of performance and reasoning quality for behavior analysis.

Minimum System Requirements

To run the 7B model locally, your system should meet these specs:

• RAM: 8GB minimum (16GB recommended)
• CPU: Modern Quad-core processor (Intel i5/Ryzen 5 or newer)
• GPU (No exceptions): NVIDIA GPU with 6GB+ VRAM or Apple Silicon (M1/M2/M3) for faster generation
• Storage: ~10GB free space

Setup Steps

1. Download Ollama from ollama.ai.
2. Install the Model: Open your terminal/command prompt and run:

   ollama pull qwen2.5:7b

3. Run Ollama Server: You must allow generic origins for the browser to access it:

   # Linux/Mac
   OLLAMA_ORIGINS="*" ollama serve
   
   # Windows PowerShell
   $env:OLLAMA_ORIGINS="*"; ollama serve

---

❓ Troubleshooting

"Module not found" or "Vite not found"

❌ Error: 'vite' is not recognized or missing module face-api.js

✅ Fix: You likely forgot to cd client. Run:

cd client
npm install
npm run dev

"AI Analysis is offline" / "Click to generate analysis" does nothing

❌ The button is disabled or says "AI Offline".

✅ Fix: Ensure Ollama is running with OLLAMA_ORIGINS="*". The app needs this to bypass CORS restrictions locally.

Webcam not working

❌ Browser blocks camera access.

✅ Fix: Check permissions in your browser address bar. Ensure no other app (Zoom/Teams) is using the camera.

---

🏗️ Architecture

ADHARA/
├── client/                    # React + Vite frontend
│   ├── src/
│   │   ├── pages/            # Child activity & Teacher dashboard
│   │   └── utils/            # Analysis modules
│   │       ├── faceAnalysis.js      # Emotion & gaze detection
│   │       ├── speechAnalysis.js    # Voice pattern analysis
│   │       └── disorderDetection.js # Clinical pattern matching
│   └── public/models/        # TensorFlow face detection models
├── data/                     # Baseline datasets
└── docs/                     # Documentation

---

🎓 Use Cases

Application	Description
Schools	Early screening for learning difficulties
Special Education	Progress monitoring and intervention planning
Research	Behavioral data collection for cognitive studies
Parents	At-home learning pattern awareness

---

👨‍💻 Team

Jithendra V Anand	Aravind Lal
Lead Developer	Contributor

---

📄 License

This project is open-source and available under the MIT License.

---

Built with ❤️ for early intervention in education

⭐ Star this repo if you find it useful! ⭐