This project captures a live video stream from an IP camera(in my case I used my phone with an app called IP camera), detects text using OCR, identifies questions, and answers them using a local LLM (Mistral 7B in GGUF format) running on CPU.

• Real-time video capture from an IP camera
• Optical Character Recognition (OCR) with Tesseract
• Automatic detection of questions in the video
• Local LLM (no internet required) for answering questions
• Live visual feedback using OpenCV

• Python 3.8+
• OpenCV
• pytesseract
• llama-cpp-python
• A GGUF LLM model like mistral-7b-instruct-v0.1.Q4_K_M.gguf
• Tesseract OCR installed (with tesseract accessible in your PATH)

git clone https://github.com/psycho237-prog/Quizbox-AI-
cd Quizbox-AI-
pip install -r requirements.txt

sudo apt update
sudo apt install tesseract-ocr

1. Update the IP camera URL in main.py:

ip_camera_url = 'http://your-camera-ip:port/video'

2. Set the correct path to your local .gguf model:

llm = Llama(model_path="path/to/your-model.gguf")

3. Run the script:

python main.py

Press q to quit the video window.

You can download the Mistral model (GGUF format) from HuggingFace:

• https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF

Place it in your project folder or adjust the path accordingly.

QUESTION: What is machine learning?
ANSWER : Machine learning is a subset of AI that enables systems to learn from data...

MIT License. Feel free to use, modify, and share.

Onana Gregoire Legrand