Datasets:
Tasks:
Image Classification
Modalities:
Image
Formats:
imagefolder
Languages:
English
Size:
1K - 10K
License:
metadata
license: mit
task_categories:
- image-classification
language:
- en
Source Reference
This dataset was sourced and prepared from the original dataset hosted on Roboflow:
🔗 https://universe.roboflow.com/checkbox-detection-ztyeq/checkbox-kwtcz-qkbid/dataset/7
Checkbox Cropped Binary Dataset
This dataset contains cropped checkbox regions extracted from document images for binary classification: checked vs unchecked.
Structure
train/ checked/ unchecked/ valid/ checked/ unchecked/ test/ checked/ unchecked/
Description
- Images are cropped checkbox regions from scanned documents
- Each image belongs to one of two classes:
checkedunchecked
- Intended for document AI and form understanding tasks
Usage
The dataset can be loaded using the Hugging Face Datasets library or accessed directly as image files.
Notes
This dataset is used in conjunction with a fine-tuned PaliGemma-based checkbox classification model.