

The MNIST database was constructed from NIST's Special Database 3 and If youĭo this kind of pre-processing, you should report it in your


Such as SVM and K-nearest neighbors), the error rate improves when theĭigits are centered by bounding box rather than center of mass. With some classification methods (particuarly template-based methods, So as to position this point at the center of the 28x28 field. the images were centered in a 28x28 imageīy computing the center of mass of the pixels, and translating the image Images contain grey levels as a result of the anti-aliasing technique usedīy the normalization algorithm. To fit in a 20x20 pixel box while preserving their aspect ratio. The original black and white (bilevel) images from NIST were size normalized Your own (very simple) program to read them. These files are not in any standard image format. Some people have asked me "my application can't open your image files". If the files you downloaded have a larger size than the above, they have been Please note that your browser may uncompress these files without telling you. It is a good database for people who want to try learning techniquesĪnd pattern recognition methods on real-world data while spending minimal The digits haveīeen size-normalized and centered in a fixed-size image. Is a subset of a larger set available from NIST. Training set of 60,000 examples, and a test set of 10,000 examples. The MNIST database of handwritten digits, available from this page, has a MNIST handwritten digit database, Yann LeCun, Corinna Cortes and Chris Burges THE MNIST DATABASE of handwritten digits Yann LeCun, Courant Institute, NYU Corinna Cortes, Google Labs, New York Christopher J.C.
