GUIrilla-Trees

Description

GUIrilla MacApp Trees is a large-scale dataset of hierarchical, accessibility-driven representations of macOS applications.

Each tree captures UI states and user interactions across full-desktop environments, providing a structured view of how applications evolve under user actions. Built using the macOS Accessibility API, these trees encode both the semantic structure of UI elements and their transitions over time.

This dataset serves as a reusable structural abstraction of desktop GUI behavior and can be used for:

• UI understanding and structured representation learning
• Retrieval and search over application states
• Automated UI testing and analysis
• Training and evaluation of desktop agents

Unlike screenshot-only datasets, GUIrilla-Trees exposes the underlying accessibility hierarchy, enabling more precise and semantically grounded modeling of user interfaces.

Repository layout & loading

⚠️ The Hub dataset viewer is intentionally disabled: the data ships as 54 large .tar.gz archives (~650 GB total) that the viewer cannot preview.

Each archive under graphs/ is one GUIrilla crawl session, named by its start time (YYYY-MM-DD_HH-MM-SS_processed.tar.gz). Inside, crawled applications are grouped per run and identified by bundle ID:

<session-timestamp>/
└── run_<N>/
    └── <bundle.id>/                       # one crawled application
        ├── <bundle.id>_last_screen.json   # final UI state of the crawl
        └── graph/                         # serialized accessibility graph of the session
            └── images/                    # full, cropped, and segmented screenshots

Archives are independent, so you can fetch a single session instead of the full dataset:

hf download macpaw-research/GUIrilla-Trees --repo-type dataset \
  --include "graphs/2025-04-25_00-49-14_processed.tar.gz" --local-dir GUIrilla-Trees

---

License

CC-BY-NC-4.0 (see LICENSE).

This dataset contains screenshots and accessibility metadata captured from third-party macOS applications; the non-commercial licence reflects that provenance.

In short (not legal advice):

• ✅ Non-commercial research — training and evaluating models, publishing papers and benchmark results, redistributing for non-commercial purposes (all with attribution).
• ❌ Commercial use — e.g. training or benchmarking models for commercial products, or redistributing the data as part of a commercial offering — is not covered by this licence.

Interested in using this dataset commercially? Contact MacPaw Research to discuss options.

Citation

@article{garkot2025guirilla,
  title={GUIrilla: A Scalable Framework for Automated Desktop UI Exploration},
  author={Garkot, Sofiya and Shamrai, Maksym and Synytsia, Ivan and Hirna, Mariya},
  journal={arXiv preprint arXiv:2510.16051},
  year={2025},
  url={https://arxiv.org/abs/2510.16051}
}