Dataset · uoft-cs
CIFAR-10: 60K Color Images
uoft-cs/cifar10
Classic dataset of 60,000 32x32 images across 10 classes for classification validation.
What's inside
CIFAR-10 consists of 60,000 32x32 color images across 10 classes, with 6,000 images per class. It contains 50,000 training and 10,000 test images; the test split has exactly 1,000 randomly selected images per class. Training data is split into five batches of 10,000 images, and testing data into one batch. Language is listed as English, modality as image, and task as image-classification.
What you can do with it
- Training and comparing CNNs and vision models
- Validating image classification pipelines
- Benchmark for small models
- Tasks
- Image classification
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 262.6K
- Likes
- 142
Models trained on it
- thearn/cifar10-resnet9Image classification 318
- nana12233/smai-a4-cifar10-mlpImage classification 225
- nana12233/smai-a4-cifar10-ioc-nnImage classification 224
- MamaPearl/nula-cifar10-robust-v0Image classification 183
- Danny-Dasilva/embeddinggemma-2-multimodal-typed-decisionZero shot classification 152
- li-ion/cifar10-mlp-384-128Image classification 135
- li-ion/cifar10-iocnn-384-128Image classification 135
- ransom32/smai-a4-cifar10-ioc-nnImage classification 113
How to load it
from datasets import load_dataset
ds = load_dataset("uoft-cs/cifar10")