README.md

---
license: apache-2.0
task_categories:
- object-detection
language:
- en
size_categories:
- 10K<n<100K
---

# PixelWeb: The First Web GUI Dataset with Pixel-Wise Labels
[https://arxiv.org/abs/2504.16419](https://arxiv.org/abs/2504.16419)

# Dataset Description

**PixelWeb-1K**: 1,000 webpages with `raw data(screenshot, element images(PNG), DOM data)` and `annotated data(mask, contour, bbox)`.

**PixelWeb-10K**: 10,000 webpages with `raw data(screenshot, element images(PNG), DOM data)` and `annotated data(mask, contour, bbox)`.

**PixelWeb-100K**: 100,000 webpages with `raw data(screenshot, element images(PNG), DOM data)` and `annotated data(mask, contour, bbox)`. The dataset has been partitioned and packaged into 10 batches.

You need to extract the tar.gz archive by:

`tar -xzvf pixelweb_1k.tar.gz`

# Document Description

    {id}-screenshot.png   # The screenshot of a webpage

    {id}-bbox.json        # The bounding box labels of a webpage, [[left,top,width,height],...]

    {id}-contour.json     # The contour labels of a webpage, [[[x1,y1,x2,y2,...],...],...]

    {id}-mask.json        # The mask labels of a webpage, [[element_id,...],...]

    {id}-class.json       # The class labels of a webpage, [axtree_label,...]

The element_id corresponds to the index of the class.