SVHN

Summary

The Street View House Numbers (SVHN) dataset is a real-world image dataset that is used for developing machine learning and object recognition algorithms. It is similar to the MNIST dataset but contains an order of magnitude more labeled data (over 600,000 digit images) and comes from a significantly harder, unsolved, real-world problem of recognizing digits and numbers in natural scene images. The dataset is obtained from house numbers in Google Street View images and is available in two formats: original images with character-level bounding boxes and MNIST-like 32-by-32 images centered around a single character.

Size

The SVHN dataset consists of 33402 training images, 13068 test images, and 202353 extra images for the full_numbers format. For the cropped_digits format, it has 73257 training images, 26032 test images, and 531131 extra images. The total number of rows in the dataset is 879,243.

Use cases

The SVHN dataset can be used for various tasks such as image classification and object detection. It can be used to train models for digit detection and image classification, where the task is to predict a correct digit in the image. The dataset has been used in the development and evaluation of machine learning models for these tasks and has a leaderboard available for image classification on the SVHN dataset.

License

The SVHN dataset is available for non-commercial use only.