CIFAR-10
Linked on 11/24/2021
A well-known image classification dataset, 10 classes, 32 x 32 pixel images.
Dataset Characteristics
Image
Subject Area
Computer Science
Associated Tasks
Classification
Feature Type
-
# Instances
60000
# Features
-
Dataset Information
For what purpose was the dataset created?
The CIFAR-10 dataset was developed for evaluation of deep generative models in 2009 and has subsequently been widely adopted as a machine learning benchmark for image classification/object recognition. It is a subset of the original Tiny Images Dataset (from MIT and NYU), with 10 classes and more reliable labels. See also the related CIFAR-100 dataset with 100 classes. More details are provided in the linked technical report by Krizhevsky from 2009.
Who funded the creation of the dataset?
Developed by researchers at the University of Toronto
What do the instances in this dataset represent?
32 x 32 pixel color images
Are there recommended data splits?
Yes, there is a standard training set of size 50,000 and a test set of size 10,000
Does the dataset contain data that might be considered sensitive in any way?
Image datasets obtained from the Web (as CIFAR-10 is) may inadvertently contain sensitive information. The larger Tiny Images dataset has now been retracted for this reason (e.g., see discussion in Peng, Mathur, Narayanan, 2021)
Was there any data preprocessing performed?
See the technical report by Krizhevsky (2009) for details on data preprocessing
Has Missing Values?
No
Introductory Paper
By Alex Krizhevsky. 2009
Published in Technical Report, Computer Science Department, University of Toronto
Reviews
There are no reviews for this dataset yet.
pip install ucimlrepo
from ucimlrepo import fetch_ucirepo # fetch dataset cifar_10 = fetch_ucirepo(id=691) # data (as pandas dataframes) X = cifar_10.data.features y = cifar_10.data.targets # metadata print(cifar_10.metadata) # variable information print(cifar_10.variables)
CIFAR-10 [Dataset]. (2009). UCI Machine Learning Repository. https://doi.org/10.24432/C5889J.
Citations/Acknowledgements
If you use this dataset, please follow the acknowledgment policy on the original dataset website.