kmnist / README.md
tanganke's picture
Convert dataset to Parquet (#1)
6b345be verified
metadata
language:
  - ja
size_categories:
  - 10K<n<100K
task_categories:
  - image-classification
dataset_info:
  config_name: kmnist
  features:
    - name: image
      dtype: image
    - name: label
      dtype:
        class_label:
          names:
            '0': 
            '1': 
            '2': 
            '3': 
            '4': 
            '5': 
            '6': 
            '7': 
            '8': 
            '9': 
  splits:
    - name: train
      num_bytes: 26807717
      num_examples: 60000
    - name: test
      num_bytes: 4478963
      num_examples: 10000
  download_size: 30674033
  dataset_size: 31286680
configs:
  - config_name: kmnist
    data_files:
      - split: train
        path: kmnist/train-*
      - split: test
        path: kmnist/test-*
    default: true

KMNIST Dataset

lassify images from the KMNIST dataset into one of the 10 classes, representing different Japanese characters.