Skip to content

Machine Learning

Dani John edited this page Jun 30, 2023 · 22 revisions

Lively can leverage Machine Learning(ML) inferencing to create dynamic wallpapers.

Types of ML operation

Depth Wallpaper

Predict depth(distance from camera) of any 2D image and generate a 3D like parallax effect.

More coming soon..

Frequently Asked Questions (FAQ)

Q. Download failed, what do I do?

First verify that internet is working and try again. The files are hosted on github, some locations or network maybe blocking them.

You can download the files directly from the browser by following the link:

https://github.com/rocksdanister/lively-ml-models/releases

Depending on your Lively setup and the machine learning category the downloaded file needs to extracted to:

Installer

C:\Users\<UserName>\AppData\Local\Lively Wallpaper\ML\<ModelName>\

Microsoft Store

%LocalAppData%\Packages\12030rocksdanister.LivelyWallpaper_97hta09mmv6hy\LocalCache\Local\Lively Wallpaper\ML\<ModelName>\

Where <ModelName> are the following:

Depth wallpaper : Midas

Create directory if it does not exist.

Q. How does Depth Wallpaper work, can I modify it?

Depth information or Depth Map is generated on the fly when creating the wallpaper. Depth Map shows the distance of object from the camera, in the following example darker region is further away from the camera and brighter near.

Based on this information the image pixels are shifted more when closer and less when further away from camera.

Lively leverages GPU shaders to shift the pixels for performance.

The generated wallpaper is a based on this project template:

https://github.com/rocksdanister/depthmap-wallpaper

As it is a standard webpage this can be further enhanced by users, for example add a visualizer using Lively System API