#ResearchDataset #MNIST #MachineLearning #ComputerVision #DeepLearning #HandwrittenDigits #ImageRecognition #Classification #DataPreprocessing #FeatureExtraction #DataAugmentation #BenchmarkDataset
The Modified National Institute of Standards and Technology database (MNIST) is a vast collection of handwritten digits. It has a 60,000-example training set and a 10,000-example test set. It is a subset of the larger NIST Special Database 3 (numbers written by US Census Bureau personnel) and Special Database 1 (digits written by high school students), both of which contain monochrome photographs of handwritten digits. In a fixed-size picture, the digits have been size-normalized and centred. The original NIST black and white (bilevel) photos were size adjusted to fit in a 20x20 pixel box while maintaining their aspect ratio.