Context [SVHN](http://ufldl.stanford.edu/housenumbers/) is a real-world image dataset. Fragments of this dataset were preprocessed: - fields of photos that do not contain digits were cut off; - the photos were formatted to the standard 32X32 size; - three color channels were converted into one channel (grayscaled); - each of the resulting images was represented as an array of numbers; - the data were converted into `.csv` files. Content - 64000 32x32 greyscaled images of number photos with 1-5 digits (represented as arrays). - 11 categories of labels (ten "digit" categories and one "empty character" category). - Information about file names in the original dataset. Acknowledgements The original data contains a notice "for non-commercial use only". Inspiration Image recognition and classification is a huge part of machine learning practice. In addition, this data is based on real photos.