Neural Network Importance Map Generation for Graphic Design
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional tools for determining the visual importance of graphic designs and data visualizations are limited as they rely on eye gaze measurements, require human annotations, and are inaccurate for bitmap images, failing to provide interactive applications for graphic designers.
Innovation Solution
Training neural networks to generate importance maps without human eye gaze measurements or annotations, using pixel-level comparisons with ground truth datasets, enabling real-time design feedback, smart thumbnails, and design retargeting for bitmap images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional tools use eye gaze measurements to determine saliency, then they can identify important regions in natural images, but they cannot accurately determine importance of portions of bitmap images of graphic design or data visualization
Solution Approach 1:
The patent changes the fundamental parameters of the training data from natural images with eye gaze measurements to graphic design bitmap images with programmatically-defined elements. This parameter change enables the neural network to learn features specific to graphic designs while maintaining pixel-level processing capability, thereby resolving the contradiction between measurement precision and adaptability to different image types.
2Measurement precision
If existing tools require human annotations for training neural networks, then they can achieve accurate saliency prediction, but they cannot operate without human eye gaze measurements
Solution Approach 1:
The system enables automated generation of training data through programmatically-defined elements in graphic designs. The neural network is trained using automatically generated ground truth from design elements rather than human annotations, achieving both high accuracy and full automation. This self-service approach eliminates dependency on human eye gaze measurements while maintaining prediction accuracy.
3Measurement precision
If neural networks are trained with pixel level comparison loss function, then they can provide accurate importance maps, but they do not offer interactive applications to assist graphic designers
Solution Approach 1:
The patent implements feedback mechanisms where the neural network's importance maps are integrated into the graphic design workflow. The system provides real-time feedback to designers by highlighting important regions and suggesting design improvements, thereby transforming the accurate but static importance maps into interactive design assistance tools that enhance ease of operation.
Data Source
AI summary
Embodiments disclosed herein describe systems, methods, and products that train one or more neural networks and execute the trained neural network across various applications. The one or more neural networks are trained to optimize a loss function comprising a pixel-level comparison between the outputs generated by the neural networks and the ground truth dataset generated from a bubble view methodology or an explicit importance maps methodology. Each of these methodologies may be more efficient than and may closely approximate the more expensive but accurate human eye gaze measurements. The embodiments herein leverage an existing process for training neural networks to generate importance maps of a plurality of graphic objects to offer interactive applications for graphics designs and data visualizations. Based on the importance maps, the computer may provide real-time design feedback, generate smart thumbnails of the graphic objects, provide recommendations for design retargeting, and extract smart color themes from the graphic objects.