Open Shelves

Open Shelves is an open source computer vision project to identify book spines.

The dataset for this project is hosted on Roboflow. You can also try all models that have been trained with this dataset on Roboflow.

You can use this dataset to train vision models to identify bookshelves.

This dataset was designed to improve the clickable bookshelves (source code) project. A key limitation of the project was the foundation models on which it depends, Grounding DINO and Segment Anything, require signifiant computational resources to run, particularly on large numbers of images. With Open Shelves, a fine-tuned model can be trained that is faster than the Grounding DINO and Segment Anything combination.