# Is there a open source implementation of "Deep Learning Based Page Layout Analyze"?

**URL:** <https://discuss.huggingface.co/t/is-there-a-open-source-implementation-of-deep-learning-based-page-layout-analyze/19557>\
**Category:** Research\
**Created:** [June 24, 2022, 2:47am UTC](https://discuss.huggingface.co/t/is-there-a-open-source-implementation-of-deep-learning-based-page-layout-analyze/19557 "2022-06-24T02:47:52Z")\
**Posts on this page:** 6\
**Page:** 1

<div class="post-metadata">

**Author:** ![neo-benjamin](https://avatars.discourse-cdn.com/v4/letter/n/f6c823/32.png) [@neo-benjamin](https://discuss.huggingface.co/u/neo-benjamin)\
**Post date:** [June 24, 2022, 2:47am UTC](https://discuss.huggingface.co/t/is-there-a-open-source-implementation-of-deep-learning-based-page-layout-analyze/19557/1 "2022-06-24T02:47:52Z")

</div>

Is there a open source implementation of “Deep Learning Based Page Layout Analyze”?

Repo: [GitHub - leonlulu/DeepLayout: Deep learning based page layout analysis](https://github.com/leonlulu/DeepLayout)  
DeepLayout: A Semantic Segmentation Approach to Page Layout Analysis

 ![image](https://us1.discourse-cdn.com/hellohellohello/original/2X/6/6204fb678d9d7cd042c9738483cd7771eaba17bd.jpeg)

Is there a model in Huggingface that could achieve the same?

```auto
@inproceedings{li2018deeplayout,
title={DeepLayout: A Semantic Segmentation Approach to Page Layout Analysis},
author={Li, Yixin and Zou, Yajun and Ma, Jinwen},
booktitle={International Conference on Intelligent Computing},
pages={266--277},
year={2018},
organization={Springer}
}

```

---

<div class="post-metadata">

**Author:** ![eugeneware](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/eugeneware/32/348_2.png) [@eugeneware](https://discuss.huggingface.co/u/eugeneware)\
**Post date:** [June 29, 2022, 9:36am UTC](https://discuss.huggingface.co/t/is-there-a-open-source-implementation-of-deep-learning-based-page-layout-analyze/19557/2 "2022-06-29T09:36:51Z")

</div>

I’m not sure about that paper, but [this library](https://layout-parser.github.io/) is very useful, and you can plug and play with different object detection models

---

<div class="post-metadata">

**Author:** ![nashid](https://avatars.discourse-cdn.com/v4/letter/n/3bc359/32.png) [@nashid](https://discuss.huggingface.co/u/nashid)\
**Post date:** [June 29, 2022, 10:20pm UTC](https://discuss.huggingface.co/t/is-there-a-open-source-implementation-of-deep-learning-based-page-layout-analyze/19557/3 "2022-06-29T22:20:53Z")

</div>

@eugeneware [Layout Parser](https://layout-parser.readthedocs.io/en/latest/index.html) library is super interesting. But looks like it is more for understanding PDF documents.

Are you aware of any specific layout parsers to understand Web layout. Lets say it can understand web elements in a web page.

Web elements could be:

- A Table
- A Drop Down Menu
- A Numbered List
- A Bulleted List
- A Radio Button  
…

---

<div class="post-metadata">

**Author:** ![eugeneware](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/eugeneware/32/348_2.png) [@eugeneware](https://discuss.huggingface.co/u/eugeneware)\
**Post date:** [June 29, 2022, 10:34pm UTC](https://discuss.huggingface.co/t/is-there-a-open-source-implementation-of-deep-learning-based-page-layout-analyze/19557/4 "2022-06-29T22:34:48Z")

</div>

I’m not personally aware - but if you can find a dataset of annotated web layouts, you could use that library to fine tune your own library. I imagine that paper you referenced would have some reference to datasets that they benchmarked on.

---

<div class="post-metadata">

**Author:** ![nashid](https://avatars.discourse-cdn.com/v4/letter/n/3bc359/32.png) [@nashid](https://discuss.huggingface.co/u/nashid)\
**Post date:** [June 29, 2022, 10:47pm UTC](https://discuss.huggingface.co/t/is-there-a-open-source-implementation-of-deep-learning-based-page-layout-analyze/19557/5 "2022-06-29T22:47:36Z")

</div>

Yeah, that makes sense. I was looking into [LayoutLM](https://huggingface.co/docs/transformers/model_doc/layoutlm)

But I do not really understand this model could be applied to my use case.

---

<div class="post-metadata">

**Author:** ![gbenson](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/gbenson/32/26320_2.png) [@gbenson](https://discuss.huggingface.co/u/gbenson)\
**Post date:** [May 21, 2024, 10:38am UTC](https://discuss.huggingface.co/t/is-there-a-open-source-implementation-of-deep-learning-based-page-layout-analyze/19557/6 "2024-05-21T10:38:33Z")

</div>

> [@eugeneware](#):
>
> …if you can find a dataset of annotated web layouts…

@neo-benjamin the WebUI project is a dataset of annotated web layouts:

- homepage: [https://uimodeling.github.io/](https://uimodeling.github.io/)
- paper: [https://dl.acm.org/doi/abs/10.1145/3544548.3581158](https://dl.acm.org/doi/abs/10.1145/3544548.3581158)
- on here: [WebUI - a biglab Collection](https://huggingface.co/collections/biglab/webui-65b97f18ca1d5f670fa3e6fe)