Dataset Card for The Street View House Numbers (SVHN)

Dataset Details

Dataset Description

SVHN is a real-world image dataset for developing machine learning and object recognition algorithms with minimal requirement on data preprocessing and formatting. It can be seen as similar in flavor to MNIST (e.g., the images are of small cropped digits), but incorporates an order of magnitude more labeled data (over 600,000 digit images) and comes from a significantly harder, unsolved, real world problem (recognizing digits and numbers in natural scene images). SVHN is obtained from house numbers in Google Street View images.

Dataset Sources

• Homepage: http://ufldl.stanford.edu/housenumbers/
• Paper: Netzer, Y., Wang, T., Coates, A., Bissacco, A., Wu, B., & Ng, A. Y. (2011, December). Reading digits in natural images with unsupervised feature learning. In NIPS workshop on deep learning and unsupervised feature learning (Vol. 2011, No. 2, p. 4).

Dataset Structure

Total images: 99,289

Classes: 10 categories

Splits:

• Train: 73,257 images

• Test: 26,032 images

Image specs: 32×32 pixels, RGB

Example Usage

Below is a quick example of how to load this dataset via the Hugging Face Datasets library.

from datasets import load_dataset  

# Load the dataset  
dataset = load_dataset("randall-lab/svhn", split="train", trust_remote_code=True)   
# dataset = load_dataset("randall-lab/svhn", split="test", trust_remote_code=True)  

# Access a sample from the dataset  
example = dataset[0]  
image = example["image"]  
label = example["label"]  

image.show()  # Display the image  
print(f"Label: {label}")

Citation

BibTeX:

@inproceedings{netzer2011reading, title={Reading digits in natural images with unsupervised feature learning}, author={Netzer, Yuval and Wang, Tao and Coates, Adam and Bissacco, Alessandro and Wu, Baolin and Ng, Andrew Y and others}, booktitle={NIPS workshop on deep learning and unsupervised feature learning}, volume={2011}, number={2}, pages={4}, year={2011}, organization={Granada} }