Parse Images to get Text

[AnimeshSinha1309]

Develop a basic model that can get details on the book just from a photograph of the front page. This is of primary use in older / self-bound books, where the task should be relatively easier, yet scaling this project with parsing index and more will depend crucially on our ability to do this. The following would be involved in the development right now.

• Get the basic PyTesseract model working.
• Get and tag a small dataset of books, get area of bounding boxes and plot for title/author/waste.
• Start combining the texts based on overlap, proximity and size to get Title/author.

[KanishAnand]

Issues in OCR

• language of book cover page
• accuracy is still not great
• especially for large text or for cursive font
• decide distance limit of merging bounding boxes.

[AnimeshSinha1309]

We have this working, any integrations are deffered until we need it I guess, certainly not in this sprint.