Skip to content
This dataset contains 207,572 books from the, Inc. marketplace.
Python Shell
Branch: master
Clone or download
Latest commit e52fd8c Mar 11, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
Task1 Fix results link Oct 31, 2017
Task2 Update Oct 31, 2017
docs local links Apr 11, 2017
scripts Easier download Mar 11, 2019
LICENSE Fix LICENSE Oct 27, 2017 add data size Mar 11, 2019

Book Cover Dataset

This dataset contains 207,572 books from the, Inc. marketplace.


Results and related papers

Task 1: Classification

A. Book Cover Image to Genre (BookCover30)

The purpose of this task is to classify the books by the cover image. The BookCover30 dataset contains 57,000 book cover images divided into 30 classes. The training set and test set is split into 90% - 10% respectively.

Technical details

Task 2: Data Mining

Data Mining (Book32)

This task is to explore the entire book database. There are 207,572 books in 32 classes. This dataset contains book cover images, title, author, and category for each respective book.

Technical details


Full Images

Due to size constraints, the full images aren't available in this repository. However, we provide label files with URLs to the images hosted on Amazon. Note, the fidelity of the images cannot be guarenteed. A script to download them can be found in scripts.

(224 x 224 x 3) Images

Resized images for the BookCover30 dataset are available in this download.

Download (657 MB)


Paper on arXiv

B. K. Iwana, S. T. Raza Rizvi, S. Ahmed, A. Dengel, and S. Uchida, "Judging a Book by its Cover," arXiv preprint arXiv:1610.09204 (2016).

  title={Judging a Book by its Cover},
  author={Iwana, Brian Kenji and Raza Rizvi, Syed Tahseen and Ahmed, Sheraz and Dengel, Andreas and Uchida, Seiichi},
  journal={arXiv preprint arXiv:1610.09204},



All book cover images are hosted by and copyright, Inc. The the use of the book cover images is fair use for academic purposes.

You can’t perform that action at this time.