📚 Bookshelf AI Link to heading

Bookshelf AI An intelligent book discovery tool that reads your bookshelf from a photo and recommends new books to read.

Point it at a photo of your bookshelf — it extracts the titles using OCR, validates them against Open Library, and generates personalised recommendations based on your collection. Everything runs locally with no API keys required.

✨ Features Link to heading

  • 📸 OCR extraction — PaddleOCR scans the photo and lifts raw text from every visible book spine
  • 🔍 Title cleaning — Noise, fragments, and split lines are filtered and merged into clean candidate titles
  • 🗄️ Open Library match — Each title is validated against Open Library to get the canonical version and metadata
  • 🌍 Language detection — Automatically detects the language of your collection for relevant suggestions
  • 🤖 Smart recommendations — Subjects, genres, and themes from your matched books surface new titles you’ll likely enjoy
  • 🔌 No API keys or external accounts needed

🔄 Pipeline Link to heading

Photo → OCR → Clean → Validate (Open Library) → Recommend
  01      02     03           04                     05

📂 Project Structure Link to heading

main.py           Entry point — runs the full pipeline
ocr.py            OCR engine wrapper (PaddleOCR)
matcher.py        Title cleaning and Open Library matching
recommender.py    Book recommendation engine
requirements.txt  Python dependencies
bookshelf.jpg     Sample bookshelf image for testing

🧰 Requirements Link to heading

  • Python 3.9 or higher
  • Internet connection (for Open Library API queries)
  • PaddleOCR models are downloaded automatically on first run

📦 Installation Link to heading

Clone the repository:

git clone https://github.com/rustamdurdyyev/Bookshelf-AI
cd Bookshelf-AI

(Optional) Create a virtual environment:

python -m venv venv
source venv/bin/activate    # Windows: venv\Scripts\activate

Install dependencies:

pip install -r requirements.txt

Note: PaddleOCR will download its models on the first run. This may take a minute.

🚀 Usage Link to heading

Run the pipeline on the default sample image:

python main.py

To use your own bookshelf photo, edit the image_path variable at the top of main.py:

image_path = "your_bookshelf.jpg"

🧠 How It Works Link to heading

  1. OCR — PaddleOCR scans the image and extracts raw text from book spines
  2. Title Cleaning — Noise, fragments, and split lines are filtered and merged
  3. Validation — Each candidate title is matched against Open Library to get the canonical title
  4. Recommendations — Subjects and genres from your matched books are used to find new titles you might enjoy

🔄 Supported Image Formats Link to heading

.jpg, .jpeg, .png, .webp

Works best with clear, well-lit photos where book spines are visible and upright.

📜 License Link to heading

This project is open-source. Feel free to use, modify, and share.