From Photo to Depth Map, Entirely On-Device

A photo shows what a surface looks like, but not how deep it is. For my 3D PBR materials I use depth maps as displacement maps: a flat photo of a stone relief becomes a material with real surface depth. I needed a quick way to get those depth maps, so I built Depth Mapper.

Depth Mapper calculates depth maps directly on iPhone, iPad, Mac and Apple Vision Pro. A Core ML model (Depth Anything V2) runs locally, so your photos never leave your device. No account, no upload, no subscription.

The part that took me longest: bias

Depth models don't treat the top and the bottom of an image equally. You notice it quickly on reliefs, façades and textures. So instead of trusting a single estimate, Depth Mapper estimates depth several times and merges the results:

• Flip Averaging averages four orientations of the image to remove the top/bottom bias.
• Vertical Tiling splits the image into 2–6 tiles with adjustable overlap and blends them back together.
• Combined applies both, optionally with ±90° rotation variants. It is the slowest option and gives the cleanest result.

A side-by-side comparison of “Standard” and “Corrected” shows right away whether the extra processing is worth it for a given photo.

See depth, don't just compute it

B&W is ideal for 3D tools. False Color, Inferno and Turbo make small depth differences visible. You can switch the color map at any time without recalculating the depth, and a Far-to-Near legend shows what the colors mean.

Stay organized

Every depth map is saved as a session with its source image and settings. Group related work into collections, then save or share the source image and the depth map straight into your other tools. Zoom into the result to check the detail before you export.

What would you use it for?

Depth maps make great displacement and height data for 3D work. I would love to hear what you would use them for: 3D, AR, photo editing, or something I haven't thought of.