This week, I have been working on Open Shelves (source code), a project that lets you photograph a bookshelf and retrieve a list of the book titles. This project was inspired by my blog post last week on how I made my bookshelves clickable. The approach I outlined in that post had one big limitation: computation requirements.
Because a GPU was almost necessary to ensure images could be processed in a reasonable time frame in a hosted web app using the chosen computer vision models, it would have been prohibitively expensive for me to have a long-running hosted version of the application. This sparked the idea for Open Shelves.
1 point