Cartographic Relief Shading with Neural Networks.

Area of Science:

• Cartography
• Computer Science
• Artificial Intelligence

Background:

• Shaded relief effectively visualizes terrain on topographic maps.
• Digital shading algorithms struggle to match the expressiveness of hand-crafted relief art.
• Manual relief shading is a laborious process requiring specialized cartographic skills.

Purpose of the Study:

• To replicate hand-drawn relief shading using deep neural networks.
• To automate the creation of high-quality shaded relief maps.
• To capture and apply essential design principles of manual relief shading.

Main Methods:

• Training U-Net neural networks with manual shaded relief images and corresponding terrain models.
• Utilizing digital elevation models as input for the trained networks.
• Evaluating generated shaded relief based on expert assessment.

Main Results:

• Neural networks successfully replicated hand-drawn shaded relief art.
• Generated shadings closely resemble the quality and expressiveness of manual work.
• The networks learned to adapt illumination direction and vary brightness to accentuate terrain features.

Conclusions:

• Deep neural networks can effectively automate the creation of high-quality shaded relief.
• AI-generated relief shading captures key cartographic design principles.
• This method offers a fast and efficient alternative to traditional manual relief shading.