DeepBump is a machine-learning driven tool to generate normal & height maps from single pictures. See this blog post for an introduction.

Blender add-on: HugoTini/DeepBump

generates normal & height maps from single pictures. – Blender 3.3/M1 mac ok – Cycles for displacement mapping GitHub

https://github.com/HugoTini/DeepB

Installation

1. Download DeepBump as a ZIP.
2. In Blender, go to Edit -> Preferences -> Add-ons -> Install and select the downloaded ZIP file. Then enable the add-on.
3. In the add-on preference, click the ‘Install dependencies‘ button (this requires an internet connection and might take a while).

By installing those dependencies, be aware of Microsoft conditions. This add-on use available APIs to disable telemetry.

Usage

Check the blog post first video.

In the Shader Editor, in the right panel under the DeepBump tab:

Color (albedo) → Normals : Select a color image node and click Generate Normal Map.

Normals → Height (displacement) : Select the generated normal map image node and click Generate Height Map.

Normals → Curvature : Select the generated normal map image node and click Generate Curvature Map.

License

This repo is under the GPL license. The training code is currently not available.