---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: labels
    sequence:
      sequence: string
  - name: contentBoxes
    sequence:
      sequence: float64
  - name: paddingBoxes
    sequence:
      sequence: float64
  - name: borderBoxes
    sequence:
      sequence: float64
  - name: marginBoxes
    sequence:
      sequence: float64
  - name: key_name
    dtype: string
  splits:
  - name: train
    num_bytes: 1865221115.665
    num_examples: 38411
  download_size: 1501188240
  dataset_size: 1865221115.665
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

This is a repacked version of a split of the WebUI dataset into the HuggingFace datasets format. This repacked version focuses on the web element locations/labels and does not contain all data in the original dataset (e.g., element styles and full source code). Please see the original page for this data and more information about the dataset, including a related publication and copyright/license information.

https://huggingface.co/datasets/biglab/webui-7kbal

```
from datasets import load_dataset

dataset = load_dataset("biglab/webui-7kbal-elements")
```

NOTE: this dataset should undergo additional processing for use with object detection. Depending on the key_name, different multipliers should be applied to the coordinates. See the [GitHub](https://github.com/js0nwu/webui/blob/a66277390db23f9c8baaa21d8189ab499339db7c/models/screenrecognition/ui_datasets.py#L460) for more details.