verbrannter's picture
Update README.md
f016983 verified
metadata
dataset_info:
  features:
    - name: image
      dtype: image
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 1128255981.2766218
      num_examples: 330396
    - name: validation
      num_bytes: 140901068.51800725
      num_examples: 41292
    - name: test
      num_bytes: 141173556.41573605
      num_examples: 41300
  download_size: 1217963087
  dataset_size: 1410330606.210365
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*

The dataset comprises over four hundred thousand handwritten names obtained from charitable initiatives. Character Recognition employs image processing techniques to transform characters present on scanned documents into digital formats. It generally exhibits good performance with machine-printed fonts. Nonetheless, machines still encounter formidable obstacles in accurately identifying handwritten characters due to the vast diversity in individual writing styles.

The total number of first names was 206,799, while the total number of surnames was 207,024. The data was partitioned into a training set (330,396 samples), testing set (41,300 samples), and validation set (41,292 samples) respectively.

The DATASET has been cleaned from empty or partly empty entries.

I am not the owner of this dataset. I took this dataset from kaggle and transformed it to a huggingface dataset to make it easier to work with.