Skip to content

Dataset · clane9

Lightweight ImageNet version on 100 random classes

clane9/imagenet-100

Compact image classification dataset with reduced resolution for fast experiments.

What's inside

A sample of 100 randomly selected classes from the original ImageNet-1k dataset, sized between 100k and 1M examples. All RGB images are resized to 160 pixels on the shorter side and provided with a numerical class label. The dataset is linked to a research paper on contrastive representation learning (CMC).

What you can do with it

  • Fast prototyping and debugging of computer vision architectures without the cost of full ImageNet
  • Training and evaluating self-supervised and contrastive learning algorithms
  • Training lightweight image classification models from scratch or fine-tuning
Tasks
Image classification
Size
100K–1M rows
Downloads this month
10.9K
Likes
14

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("clane9/imagenet-100")