Learning Visual Importance for Graphic Designs and Data Visualizations

Explore a 20-minute conference talk from the ACM Symposium on User Interface Software and Technology that delves into automated models predicting the relative importance of elements in data visualizations and graphic designs. Discover how neural networks trained on human clicks and importance annotations can be used for effective summarization, design retargeting, and thumbnailing. Learn about the collection of a new crowdsourced importance dataset and the analysis of model predictions against ground truth importance and human eye movements. Gain insights into how these importance predictions can be integrated into interactive design tools, offering immediate feedback during the design process. Examine the comparison of importance-driven applications with current state-of-the-art methods, including natural image saliency, through user studies involving hundreds of MTurk participants.

Intro
What should visual importance look like
Related work
Eye fixations
Importance Maps
Model Details
Results
Graphic Designs
Design Variance