ff856/HCA-UI-Dataset

HCA-UI Dataset

A Rico-derived dataset prepared for research on hierarchy-aware mobile user interface (UI) generation and prototyping. This dataset accompanies the project HCA-UI and the manuscript Hierarchy-Constrained Attention for Diffusion-Based Mobile UI Prototyping.

Overview

The dataset is based on the public Rico dataset. The HCA-UI preparation process includes:

• Selection and organization of mobile UI examples from Rico.
• Natural-language descriptions associated with the selected interfaces.
• Interface categorization into finance, list, media, social, utility, and web-style UI groups.
• Component structure information, including UI hierarchies and component geometry derived from the source data.

The companion paper reports experiments using 57,600 Rico training interfaces and 6,400 Rico test interfaces. These are the experimental counts reported in the manuscript; refer to the included files and the paper for the applicable data splits and processing details.

Download

The files are distributed as a multi-volume 7-Zip archive to make a large collection easier to upload and share. The archive consists of numbered parts ending in .7z.001, .7z.002, and so on. All parts are required to extract the complete dataset.

Option A: Download from the website

1. Open the Files and versions tab.
2. Download every numbered .7z.* part to the same local folder.
3. Install 7-Zip if necessary.
4. Right-click the .7z.001 part and choose 7-Zip → Extract Here (or Extract to...).

Do not try to extract .002, .003, etc. separately. Their contents are joined automatically when extraction starts from .001.

Option B: Download from the command line

Install the Hugging Face CLI:

pip install -U huggingface_hub

Download the repository files:

hf download ff856/HCA-UI-Dataset --repo-type dataset --local-dir ./HCA-UI-Dataset

Then run 7-Zip on the first archive part. For example, if the first part is named HCA-UI-Dataset.7z.001:

7z x HCA-UI-Dataset.7z.001

Replace that example filename with the actual .7z.001 filename shown in the repository. Depending on compression, the number of parts may differ from six; download all parts that are present.

Data and Research Use

The collection is intended for research in:

• Text- and structure-conditioned mobile UI image generation.
• Hierarchy-aware component layout modeling.
• Evaluation of component relationships and visual structure.
• Structured UI prototyping and editing.

The text descriptions, category assignments, and organized examples are intended to support correspondence between UI-level descriptions and component hierarchies.

Provenance and Rights

Original source: Rico — Interaction Mining.

Derived material: Selection, organization, interface categorization, and additional natural-language descriptions for the HCA-UI research project.

Important: This repository includes material derived from Rico, including original UI screenshots and/or structural metadata. Public accessibility does not establish a right to redistribute or reuse all original assets. No blanket license for the original Rico content is asserted by this dataset card. Users should review the original dataset's terms and any applicable third-party rights before redistributing or using source assets. Rights in the added annotations do not supersede rights in the original content.

Related Resources

• Code: https://github.com/ffdsj/HCA-UI
• Dataset: https://huggingface.co/datasets/ff856/HCA-UI-Dataset
• Original Rico dataset: https://www.interactionmining.org/archive/rico
• Companion manuscript: Hierarchy-Constrained Attention for Diffusion-Based Mobile UI Prototyping, Kun Xu, Shang Zheng, Hualong Yu, and Boyin Jin.

Citation

If this dataset is useful for your work, please cite the original Rico dataset and the HCA-UI companion manuscript. A formal publication identifier for the companion manuscript can be added here when available.