Sketch2aia (Mobile User Interface Sketches)
| Description | Dataset of 374 photos of hand-drawn sketches of App Inventor apps used for development of the Sketch2aia model for automatic generation of App Inventor wireframes from hand-drawn sketches. |
| Data collection | This dataset is composed of photos of hand-drawn user interface Sketches of App Inventor apps, made either during the development of App Inventor apps or by volunteers, based on screenshots of App Inventor apps. |
| Data formats | Training (/train):237 images in JPG (.jpg) format with 720×1280 pixels, each accompanied by a JSON (.json) file with manually attributed bounding box annotation for 10 different classes of UI elements (Screen, Label, Button, Switch, Slider, TextBox, CheckBox, ListPicker, Image and Map), used to train the Sketch2aia model.
Validation (/val):42 images in JPG (.jpg) format with 720×1280 pixels, each accompanied by a JSON (.json) file with manually attributed bounding box annotation for 10 different classes of UI elements (Screen, Label, Button, Switch, Slider, TextBox, CheckBox, ListPicker, Image and Map), used to test the Sketch2aia model. Additional Images (/not_used):95 images in JPG (.jpg) format with 720×1280 pixels. Some images are accompanied by a JSON (.json) file with manually attributed bounding box annotation for 10 different classes of UI elements (Screen, Label, Button, Switch, Slider, TextBox, CheckBox, ListPicker, Image and Map), while others have not yet been labeled. This portion of the dataset was collected during user evaluation of the Sketch2aia model, and have not been directly used to train or test the object detection model. |
| Download | Full dataset, organized into three main folders: train, val and not_used Git with the code of our model |
| Licence | Our research is available under creative commons Attribution-NonCommercial-ShareAlike 4.0 International |
| Citation | de Souza Baulé, D., Gresse von Wangenheim, C. and von Wangenheim, A. (2019). Hand-drawn User Interface Sketch dataset for the automatic generation of App Inventor Wireframes with Deep Learning, GQS/INCoD/INE/UFSC [Data set]. |
| Keywords | Sketch2aia, App Inventor, Mobile application, Android, Deep learning, Wireframe |