StudySphere

A real-time classroom comprehension analysis platform built for educators. StudySphere uses computer vision and deep learning to analyze student facial expressions during live sessions — giving teachers instant, data-driven insights into how well their class is following along.

---

Screenshots

Login Screen	Class Setup
Dashboard	Session Report

---

Features

• Real-Time Video Monitoring — Live webcam feed with per-face comprehension labels updated every frame
• Snapshot Evaluation — Upload a classroom photo for instant comprehension analysis across all detected faces
• Session Reports — Auto-generated reports with comprehension timeline logged every 30 seconds, concentration trend graph, and final verdict
• Multi-Theme UI — Switch between Lilac, Yellow, and Blue themes

---

How It Works

1. Faces are detected in each webcam frame using OpenCV Haar Cascades
2. Each face crop is passed to a MobileNetV2 model fine-tuned on FER2013
3. The model outputs a confidence score mapped to:
   • 🟢 Understood (≥ 0.65)
   • 🟡 Uncertain (0.40 – 0.65)
   • 🔴 Not Understood (< 0.40)
4. Scores are aggregated using a weighted formula grounded in Cognitive Load Theory
5. Every 30 seconds, a window report is logged and pushed to the frontend dashboard

---

Tech Stack

Layer	Technologies
Frontend	React, Vite, Recharts, Custom CSS
Backend	Python, Flask, flask-cors
ML & Vision	TensorFlow/Keras, OpenCV, MobileNetV2
Dataset	FER2013 (binary: Understood / Not Understood)

---

Model

The trained model is not included in this repo due to file size. Download it here: studysphere_model.keras

After downloading, place it at:

backend/studysphere_model.keras

---

Model Performance

Metric	Value
Test Accuracy	66.94%
Test Loss	0.6065
Validation Accuracy	66.21%

The model is a MobileNetV2 fine-tuned on a binary-restructured FER2013 dataset (Understood / Not Understood). Accuracy reflects the challenge of real-world classroom facial expression analysis under varying lighting, angles, and occlusions.

---

Getting Started

Prerequisites

• Python 3.10 (conda cnn environment)
• Node.js 18+
• Webcam access

1. Clone the repo

git clone https://github.com/parinaB/StudySphere.git
cd studysphere

2. Start the backend

conda activate cnn
cd backend
python app.py

Backend runs at http://localhost:5050

3. Start the frontend

cd frontend
npm install
npm run dev

Frontend runs at http://localhost:5173

---

📡 API Endpoints

Method	Endpoint	Description
POST	/evaluate	Upload image, returns comprehension breakdown
GET	/stream	Live MJPEG webcam stream
GET	/comprehension	Latest real-time comprehension score
POST	/start-video	Start session processing thread
POST	/stop-video	Stop session cleanly
GET	/session-report	Full 30-sec timeline + final verdict

---

👩‍💻 Author

Built by Parina — ML student, edutech researcher.