By: Blogs on jonathanBieler.github.io
Re-posted from: https://jonathanbieler.github.io/blog/focal.jl/
Models for estimating depth from a single image
(e.g, Depth-Anything-V2,
MiDaS) are pretty good nowadays, to the point
where they can be used for photo editing (even Photoshop use them).
In this article, I explore some applications of a depth mask for phography using Depth-Anything-V2 in Julia. The most obvious use case is to simulate bokeh (especially for smartphone that often can’t produce shallow depth of field optically), but a depth mask can also be used to add or remove haze from an image, simulate lighting with a flash or provide a natural vignette. These can help to add depth to an image and emphasize its subject.
