LectroNotes 🎓

An AI-powered lecture assistant that transforms audio and video lectures into structured study materials automatically.

What It Does

Upload any lecture audio or video and get:

• 📝 Transcript — Full text transcription using OpenAI Whisper
• 📋 Summary — Concise summary using Facebook BART
• 🔑 Key Concepts — Important points extracted using Sentence-BERT
• 📊 Topic Clusters — Content grouped by topics using KMeans
• ⭐ Quality Score — Lecture quality rated using a custom TensorFlow ANN
• ❓ Quiz — Auto-generated MCQs from key concepts
• 🃏 Flashcards — Study cards for quick revision

Tech Stack

Layer	Technology
Speech to Text	OpenAI Whisper
Summarization	Facebook BART Large CNN
Key Concepts	Sentence-BERT + Cosine Similarity
Topic Clustering	KMeans + PCA
Quality Scoring	TensorFlow/Keras ANN
API	FastAPI
Frontend	React.js

Project Structure

LectroNotes/ ├── backend/ │ ├── api/ │ │ └── notes.py # API endpoints │ ├── services/ │ │ ├── transcriber.py # Whisper transcription │ │ ├── summarizer.py # BART summarization │ │ ├── extractor.py # Key concept extraction │ │ ├── quality_scorer.py # TensorFlow quality scoring │ │ └── quiz_generator.py # Quiz and flashcard generation │ └── models/ # Trained model files (not included) ├── frontend/ # React application ├── main.py # App entry point └── train.py # Model training script

How to Run

1. Install Backend Dependencies

pip install fastapi uvicorn whisper transformers torch sentence-transformers tensorflow scikit-learn pandas numpy static-ffmpeg

2. Train the Quality Scoring Model

python train.py

3. Start the Backend

python main.py

4. Start the Frontend

cd frontend
npm install
npm start

How It Works

Audio/Video Upload ↓ Whisper → Transcript ↓ BART → Summary ↓ Sentence-BERT → Key Concepts ↓ KMeans → Topic Clusters ↓ TensorFlow ANN → Quality Score ↓ Quiz + Flashcards Generated