README.md

metadata

dataset_info:
  features:
    - name: image
      dtype: image
    - name: ground_truth
      struct:
        - name: bbox
          list:
            - name: h
              dtype: float64
            - name: id
              dtype: string
            - name: w
              dtype: float64
            - name: x
              dtype: float64
            - name: 'y'
              dtype: float64
        - name: dateOfBirth
          dtype: string
        - name: fullName
          dtype: string
        - name: gender
          dtype: string
        - name: id
          dtype: string
        - name: label_dateOfBirth
          dtype: string
        - name: label_fullName
          dtype: string
        - name: label_gender
          dtype: string
  splits:
    - name: train
      num_bytes: 1172837507
      num_examples: 10000
  download_size: 1168963461
  dataset_size: 1172837507
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - image-to-text
language:
  - km
pretty_name: SynthKhmer-10k
size_categories:
  - 10K<n<100K

SynthKhmer 10k

Image Size: 896x672

Dataset Creation

• Background images are from https://picsum.photos/
• Profile Photos are from https://thispersondoesnotexist.com/
• Names are sampled from Khmer Dictionary