Powerful machine learning models are able to read and locate characters in images and videos. These models can be used in a wide array of tasks. Ranging from automatic number-plate recognition, digitizing archives to blurring out sensitive text in images and assisting blind and visually impaired persons.
The Ghent University Library has a collection of scanned records of 1.1 million catalogue cards. They already had partially digitized some records but a big collection of books remained unavailable in the main database. They risked having as much as a complete floor in our book depository full of items which are not used because these cannot be discovered on the web.
By performing OCR with the Google Vision API and by running entity detection, we were able to turn the scanned cards into valuable records that could be added to the main library database. On top of that some cleaning was done to delete outdated records, remove duplicates and extract the call number of each card (which indicates the location of a book). Thereby making over 1 million forgotten books available to the public again.
Read our blog post to get more info!