Towards Automated Infographic Design: Deep Learning-based Auto-Extraction of Extensible Timeline

Area of Science:

• Computer Vision
• Artificial Intelligence
• Graphic Design Automation

Background:

• Infographic design requires balancing perceptual effectiveness and visual style, a complex task for both experts and non-experts.
• The demand for automated infographic design tools is growing, particularly for widely used formats like timeline infographics.

Purpose of the Study:

• To develop an end-to-end approach for automatically extracting extensible timeline templates from bitmap images.
• To address the challenges in manual infographic design by automating the template extraction process.

Main Methods:

• A deconstruction and reconstruction paradigm was employed.
• A multi-task deep neural network was utilized for deconstruction, parsing global and local information from timeline images.
• A reconstruction pipeline including Non-Maximum Merging, Redundancy Recover, and DL GrabCut was used to extract templates.

Main Results:

• The approach demonstrated effective extraction of extensible templates from both synthesized and real-world timeline infographics.
• Quantitative evaluations on datasets of 4296 synthesized and 393 real-world images confirmed the method's effectiveness.
• Qualitative results showed successfully extracted templates and generated timelines, validating the approach's performance.

Conclusions:

• The proposed method can effectively extract extensible templates from existing timeline infographics.
• This contributes a significant step towards automated infographic design, particularly for timeline visualizations.