LoFi Sketch: A Large Scale Dataset of Smartphone Low Fidelity Sketches

Vinoth Pandian Sermuga Pandian, Abdullah Shams, Sarah Suleri, Prof. Dr. Matthias Jarke
2022 CHI Conference on Human Factors in Computing Systems Extended Abstracts  
Recent research on automating the transformation of low fdelity (LoFi) sketches to code using Deep Neural Networks require a largescale dataset for generalizable results. This paper introduces the LoFi sketch dataset, an open-access dataset of hand-drawn smartphone LoFi sketches, to facilitate this research domain. It contains 4,527 LoFi sketches annotated with 21 categories of UI elements. Through annotations, it provides the category, position, and location of 41,560 constituent UI elements.
more » ... e collected these sketches from 361 participants from 76 countries through questionnaires and by acquiring premade LoFi sketches to ensure the dataset's diversity and ecological validity. These sketches were drawn using pen, pencil, mouse, touch, and stylus as the input medium and available in raster and vector format. This dataset enables further research on several avenues of AI assistance in UI design, such as UI element detection, sketch completion, UI layout refnement, and Sketch-based image retrieval. CCS CONCEPTS • Human-centered computing → Interface design prototyping; User interface design; Wireframes; • Computing methodologies → Image representations.
doi:10.1145/3491101.3519624 fatcat:i5zkska27nafrfbdricjrskhmu