The CIFAR-10 database (Canadian Institute For Advanced Research database) is a large collection of natural color images. It has a training set of 50,000 examples, and a test set of 10,000 examples. It is a subset of a larger 80 million tiny images dataset which contains 32x32 color images. The images have been categorized into 10 mutually exclusive classes. The original images were collected from the web and labeled by human annotators. The images are 32x32 pixels with 3 color channels (RGB). The dataset consists of 60,000 32x32 color images in 10 different classes, with 6,000 images per class. The 10 different classes represent airplanes, cars, birds, cats, deer, dogs, frogs, horses, ships, and trucks. There are 5,000 training images and 1,000 test images per class. The classes are completely mutually exclusive; there is no overlap between automobiles and trucks. "Automobile" includes sedans, SUVs, and things of that sort. "Truck" includes only big trucks. Neither includes pickup trucks.
Variants: CIFAR-10 (20% subset), Superpixel CIFAR10, Split CIFAR-10 (5 tasks), split-cifar10, NATS-Bench Size, CIFAR-10, Leave-One-Class-Out CIFAR-10, cifar10_quality_drift, CIFAR-10 (partial ratio 0.5), CIFAR-10 (partial ratio 0.3), CIFAR-10 (partial ratio 0.1), CIFAR-10, Human Noise, CIFAR-10, 60% Symmetric Noise, CIFAR-10, 60% IDN, Cifar10 (5 tasks), CIFAR-10, 40% Symmetric Noise, CIFAR-10,40 Labels, CIFAR-10, 40% IDN, CIFAR-10, 4000 Labeled Samples, CIFAR-10, 30 Labels, CIFAR-10, 30% Asymmetric Noise, CIFAR-10, 20% IDN, CIFAR-10 (20% data), CIFAR-10, 20% Asymmetric Noise, CIFAR-10, 2000 Labeled Samples, cifar-10, 10 Labels, CIFAR-10 (10% data), CIFAR-10, 100 Labels, CIFAR10 (10,000), cifar10, CIFAR10_CATS, CIFAR-10 (with noisy labels), noise padded CIFAR-10, cifar10, 10 labels, cifar-10,4000, CIFAR10 100k, CIFAR-10 model detecting CIFAR-10, CIFAR-10 image generation, CIFAR-10 WRN-28-10 - 200 Epochs, CIFAR-10-LT (ρ=100), CIFAR-10-LT (ρ=10), CIFAR-10, 80 Labels, CIFAR-10, 500 Labels, CIFAR-10, 20 Labels, CIFAR-10 vs CIFAR-100, CIFAR-10 ResNet-18 - 200 Epochs, CIFAR-10 (Conditional), cifar10, 250 Labels, One-class CIFAR-10, CIFAR-10, 4000 Labels, CIFAR-10, 40 Labels, CIFAR-10, 250 Labels, CIFAR-10, 2000 Labels, CIFAR-10, 1000 Labels, CIFAR-10 Image Classification, CIFAR-10
This dataset is used in 27 benchmarks:
Recent papers with results on this dataset: