Dataset Overview

Feature Information

• image: Image data, data type is image.
• instruction: Instruction text, data type is string.
• bbox: Bounding box, data type is a sequence of floats.
• resolution: Resolution, data type is a sequence of integers.
• source: Data source, data type is string.
• platform: Platform information, data type is string.
• name: Name, data type is string.
• description: Description, data type is string.
• type: Type, data type is string.
• OCR: OCR text, data type is string.
• language: Language, data type is string.
• purpose: Purpose, data type is string.
• expectation: Expectation, data type is string.

Data Split

• train: Training set, containing 63,530 samples, size 27,758,382,015.75 bytes.
• validation: Validation set, containing 7,944 samples, size 3,476,872,245.0 bytes.
• test: Test set, containing 7,938 samples, size 3,297,859,832.75 bytes.

Dataset Size

• Download size: 3,400,799,177 bytes
• Total dataset size: 34,533,114,093.5 bytes

Configuration Information

• config_name: default
  • data_files:
    • train: path is data/train-*
    • validation: path is data/validation-*
    • test: path is data/test-*