ZoeDepth

Name: Zoe Depth Map
Author: Intel Labs

byIntel Labs

Extract high-precision metric depth maps with crisp spatial boundaries and linear distance scaling for 3D displacement and ControlNet workflows

Extract Depth Map

Grayscale depth map of five cats seen from behind, tails raised, glowing white nearest the camera

How Zoe Depth Map works

From standard photograph to metric depth map in three frictionless steps.

Zoe Depth Map node with a photo of an empty opera stage and red rowboat wired in, showing its grayscale depth map

Upload source image

Select an input photograph with defined spatial depth, layered subjects, or architectural perspective.

Zoe Depth Map node selected, its Image input holding the opera-stage photo above the grayscale depth output

Process metric depth

The transformer backbone dynamically routes indoor and outdoor cues to compute millimeter-accurate linear distance.

Zoe Depth Map node output: grayscale depth map of tall stage flats, the floor brightening toward the camera

Export depth pass

Download the high-contrast depth map for ControlNet conditioning, 3D mesh displacement, or spatial VFX.

What Zoe Depth Map is good at

Explore how hybrid relative-metric estimation and automated scene routing deliver unrivaled spatial fidelity.

Grayscale depth map of a night street, a figure bright in the foreground, parked cars and people fading behind

Relative and Metric Fusion

Combines metric distance accuracy with relative depth generalization, preserving true physical proportions across layered spatial planes without geometric distortion.

Grayscale depth map of a woman with crossed arms and two layered paper fans, arms and fans bright in front of the body

Complex Overlap Resolution

Isolates tightly overlapping foreground elements such as crossed arms, splayed fingers, and layered props with razor-sharp volumetric boundaries.

Grayscale depth map of a person in a blazer standing out bright white against a dark street

Intelligent Scene Routing

Automatically routes inputs to dedicated indoor (NYU Depth v2) or outdoor (KITTI) metric heads based on latent scene classification without manual intervention.

Grayscale depth map of a terraced stone plaza with steps and trees, nearer levels lighter

3D Displacement Precision

Generates high-contrast linear grayscale depth values that map distance directly to physical height fields for Blender, Unreal Engine, and Maya.

Made with Zoe Depth Map

A visual gallery demonstrating crisp geometric plane isolation, complex limb depth resolution, and surface displacement across diverse subjects.

Grayscale depth map of a toy pony standing on a floor that brightens toward the camera

Granular grayscale depth data ready for complex objects and scenes

Grayscale depth map of a blank signboard on a single post, bright against a black background

Distance-accurate depth plates ready for 3D generation

Grayscale depth map of a box with a curved tube resting on top

Multiple object interaction and scene depth mapping

$Grayscale depth map of a fractured stoneware vessel with a broken-open front and a pile of shards on a plinth$

Fractured ceramic sculpture with tactile surface detail

Grayscale depth map of a woman sitting cross-legged on wide steps beside a dark doorway

Seated figure lifted cleanly off the steps behind her

What people build with Zoe Depth Map

See how 3D generalists, VFX compositors, and AI creators harness metric depth for precise spatial conditioning.

Architectural spatial reconstruction

Translate flat 2D architectural photographs into geometrically accurate 3D scene geometry, retaining exact floor-to-ceiling planes and window recess depths.

Displacement mapping for 3D meshes

Convert photorealistic surface photographs into displacement and height maps for Blender, Unreal Engine, and Maya with clean linear depth falloff.

Pose and spatial ControlNet guidance

Supply downstream generative models with clean spatial maps that preserve complex human limb intersections and hand foreshortening without prompt bleed.

Z-depth passes for post-production

Generate instant Z-depth channels from live-action footage stills to drive realistic lens blur, atmospheric haze, and particle layering in compositing suites.

E-commerce product depth extraction

Isolate product layers and extract physical volume contours for interactive 3D web viewers and augmented reality product previews.

Zoe Depth Map specs

Plans and credits

Inputs: Image
Output: image
Cost: 10 credits per run

Frequently asked questions

Yes, Zoe is the common shorthand for ZoeDepth (Zero-shot Transfer by Combining Relative and Metric Depth), a monocular depth estimation architecture developed by Intel Labs and KAUST.

ZoeDepth fuses metric depth estimation with relative depth generalization. Unlike older preprocessors that generate soft, ambiguous depth blobs, ZoeDepth dynamically routes images to indoor or outdoor metric heads to compute linear, physically accurate distance values along the Z-axis.

ZoeDepth is ideal for images with distinct foreground-to-background separation, deep interior or exterior architectural perspective, and complex overlapping limbs or hands. It is widely used to create 3D displacement maps in software like Blender and precise ControlNet reference guides.

Avoid ZoeDepth on flat 2D illustrations, unshaded vector graphics, and extreme fish-eye lens photos, as these lack the perspective cues required for metric depth prediction. For simple background segmentation or high-speed processing, lighter alternatives like LeReS are better suited.

ZoeDepth was created by researchers at Intel Labs and KAUST and initially released in February 2023. It introduced a Dense Prediction Transformer architecture with a BEiT backbone and metric bins module for zero-shot metric depth transfer.

Try Zoe Depth Map on Fuser

Extract high-precision metric depth maps with crisp spatial boundaries and linear distance scaling for 3D displacement and ControlNet workflows

Extract Depth Map

Zoe Depth Map docs All models

Gemini Image

Google DeepMind

GPT Image

OpenAI

Seedream

ByteDance (Seed team)

FLUX.3 Image

Black Forest Labs

FLUX.2