Real Bookshelf to Virtual Bookshelf 📚

The idea is to take an image of a real life bookshelf and generate a digital list of books, along with a Goodreads link for each of them.

Project Structure

Final_Project_Bookshelf.ipynb: This is the main notebook where the experimentation was done and the system was implemented. This was eventually refactored to shift all the core functions to the individual .py file.
app.ipynb: This notebook is the interactive application hosted at http://bookshelf.karangupta.net.
east.py: Contains functions for detecting text in images using the EAST Text Detection model.
gocr.py: Contains functions for doing OCR using Google Cloud Vision API and for generating Google and GoodReads links for the books.
segment.py: Contains functions for segmenting an image into segments corresponding to individual books.
utils.py: Contains utility functions used across the project.
google_app_creds.default.json: Contains the credentials for calling the Google Cloud Vision API. For more info on getting the credentials, please check this link. This file should be renamed to google_app_creds.json before deployment.
requirements.txt: File used to install project dependencies.
./images: Directory that contains sample images.
./experiments: Directory that contains notebooks with various experiments done over the course of the project (may not work now).

Make sure Python3 is installed.
Install Opencv on a Debian system using: sudo apt install python3-opencv
Install the dependencies by running: pip3 install -r requirements.txt
Launch Jupyter notebook server and run the Final_Project_Bookshelf.ipynb notebook as usual.
Launch the application by running: voila app.ipynb