metadata

license: apache-2.0
size_categories:
  - 10K<n<100K
task_categories:
  - image-feature-extraction
dataset_info:
  features:
    - name: image
      dtype: image
  splits:
    - name: train
      num_bytes: 2156767982
      num_examples: 26000
  download_size: 2183967663
  dataset_size: 2156767982
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Imagenet Mini Dataset

This dataset is a subset of the Imagenet validation set containing 26,000 images. It has been curated to have equal class distributions, with 26 randomly sampled images from each class. All images have been resized to (224, 224) pixels, and are in RGB format.

Citation

If you use this dataset in your research, please cite the original Imagenet dataset:

Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009). Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition (pp. 248–255).