MNIST

E74103

benchmark dataset dataset handwritten digit dataset

MNIST is a widely used benchmark dataset of handwritten digit images commonly employed for training and evaluating image classification algorithms in machine learning and computer vision.

Statements (47)

Predicate	Object
instanceOf	benchmark dataset → dataset → handwritten digit dataset →
backgroundColor	black →
basedOn	NIST Special Database 1 → NIST Special Database 3 →
benchmarkStatus	canonical toy dataset in machine learning →
classLabels	digits 0 through 9 →
commonModelType	convolutional neural network → multilayer perceptron →
creator	Christopher J. C. Burges → Corinna Cortes → Yann LeCun →
dataSource	scanned handwritten digits →
dataType	grayscale images →
digitColor	white →
domain	computer vision → machine learning →
fileFormat	IDX →
fullName	Modified National Institute of Standards and Technology database →
hostedBy	Yann LeCun’s website →
imageChannels	1 →
imageFile	t10k-images-idx3-ubyte → train-images-idx3-ubyte →
imageHeight	28 pixels →
imageResolution	28x28 pixels →
imageWidth	28 pixels →
inspiredDataset	EMNIST → Fashion-MNIST → KMNIST →
introducedInPublication	Gradient-based learning applied to document recognition →
labelFile	t10k-labels-idx1-ubyte → train-labels-idx1-ubyte →
license	freely available for research and educational use →
numberOfClasses	10 →
preprocessingStep	centering in a fixed-size image → size normalization →
publicationYear	1998 →
task	handwritten digit recognition → image classification →
testSetSize	10000 →
totalImages	70000 →
trainingSetSize	60000 →
typicalUse	benchmarking classification algorithms → educational examples in deep learning → training neural networks →
valueRange	0 to 255 grayscale intensity →

Referenced by (4)

Subject (surface form when different)	Predicate
torchvision →	dataset
Gradient-based learning applied to document recognition →	datasetUsed
A fast learning algorithm for deep belief nets →	evaluationDataset
LeNet →	notableDataset