CIFAR100

CIFAR-100 is an image dataset used for object recognition in computer vision. It is part of the CIFAR family and was created by Alex Krizhevsky, Vinod Nair, and Geoffrey Hinton in 2009. The dataset extends CIFAR-10 by providing 100 fine classes, grouped into 20 coarse superclasses.

The dataset consists of 60,000 32x32 color images, with 50,000 training samples and 10,000 test samples. On average there are 500 training images per fine class and 100 test images per fine class, for a total of about 600 images per class. Each image is labeled with a fine label (0-99) and a coarse label (0-19). The data are distributed in batch files: five training batches of 10,000 images each and one test batch of 10,000 images. Each image is stored as a 32x32x3 array with separate label arrays and a mapping of filenames.

CIFAR-100 is widely used as a standard benchmark for image classification and is more challenging than CIFAR-10 due to the larger number of classes and fewer samples per class. It is commonly used to evaluate convolutional neural networks and other deep learning models. Preprocessing typically includes per-channel normalization and data augmentation such as random cropping and horizontal flipping during training.