GUI Odyssey

This repository is the official implementation of GUI Odyssey.

GUI Odyssey: A Comprehensive Dataset for Cross-App GUI Navigation on Mobile Devices
Quanfeng Lu, Wenqi Shao✉️⭐️, Zitao Liu, Fanqing Meng, Boxuan Li, Botong Chen, Siyuan Huang, Kaipeng Zhang, Yu Qiao, Ping Luo✉️
✉️ Wenqi Shao (shaowenqi@pjlab.org.cn) and Ping Luo (pluo@cs.hku.hk) are correponding authors.
⭐️ Wenqi Shao is project leader.

💡 News

• 2024/06/24: The data of GUI Odyssey is released! Please check out OpenGVLab/GUI-Odyssey!
• 2024/06/13: The paper of GUI Odyssey is released!

🔆 Introduction

GUI Odyssey is a comprehensive dataset for training and evaluating cross-app navigation agents. GUI Odyssey consists of 7,735 episodes from 6 mobile devices, spanning 6 types of cross-app tasks, 201 apps, and 1.4K app combos.

🛠️ Data collection pipeline

GUI Odyssey comprises six categories of navigation tasks. For each category, we construct instruction templates with items and apps selected from a predefined pool, resulting in a vast array of unique instructions for annotating GUI episodes. Human demonstrations on an Android emulator capture the metadata of each episode in a comprehensive format. After rigorous quality checks, GUI Odyssey includes 7,735 validated cross-app GUI navigation episodes.

📝 Statistics

Splits	# Episodes	# Unique Prompts	# Avg. Steps	Data location	Model
Total	7,735	7,735	15.4	GUI-Odyssey	OdysseyAgent
Train-Random & Test-Random	5,802 / 1,933	5,802 / 1,933	15.4 / 15.2	random_split.json	OdysseyAgent-Random
Train-Task & Test-Task	6,719 / 1,016	6,719 / 1,016	15.0 / 17.6	task_split.json	OdysseyAgent-Task
Train-Device & Test-Device	6,473 / 1,262	6,473 / 1,262	15.4 / 15.0	device_split.json	OdysseyAgent-Device
Train-App & Test-App	6,596 / 1,139	6,596 / 1,139	15.4 / 15.3	app_split.json	OdysseyAgent-App

💫 Dataset Access

The whole GUI Odyssey is hosted on Huggingface.

Clone the entire dataset from Huggingface:

git clone https://huggingface.co/datasets/OpenGVLab/GUI-Odyssey

And then move the cloned dataset into ./data directory. After that, the structure of ./data should look like this:

GUI-Odyssey
├── data
│   ├── annotations
│   │   └── *.json
│   ├── screenshots
│   │   └── data_*
│   │        └── *.png
│   ├── splits
│   │   ├── app_split.json
│   │   ├── device_split.json
│   │   ├── random_split.json
│   │   └── task_split.json
│   ├── format_converter.py
│   └── preprocessing.py
└── ...

Then organize the screenshots folder:

cd data
python preprocessing.py

Finally, the structure of ./data should look like this:

GUI-Odyssey
├── data
│   ├── annotations
│   │   └── *.json
│   ├── screenshots
│   │   └── *.png
│   ├── splits
│   │   ├── app_split.json
│   │   ├── device_split.json
│   │   ├── random_split.json
│   │   └── task_split.json
│   ├── format_converter.py
│   └── preprocessing.py
└── ...

⚙️ Detailed Data Information

Please refer to this.

🚀 Quick Start

Please refer to this to quick start.

📖 Release Process

• Dataset
  • Screenshots of GUI Odyssey
  • annotations of GUI Odyssey
  • split files of GUI Odyssey
• Code
  • data preprocessing code
  • inference code
• Models

🖊️ Citation

If you feel GUI Odyssey useful in your project or research, please kindly use the following BibTeX entry to cite our paper. Thanks!

@article{lu2024gui,
  title={GUI Odyssey: A Comprehensive Dataset for Cross-App GUI Navigation on Mobile Devices},
  author={Lu, Quanfeng and Shao, Wenqi and Liu, Zitao and Meng, Fanqing and Li, Boxuan and Chen, Botong and Huang, Siyuan and Zhang, Kaipeng and Qiao, Yu and Luo, Ping},
  journal={arXiv preprint arXiv:2406.08451},
  year={2024}
}