- Image
- Analyze
Depth Anything
Purpose
Section titled “Purpose”Depth Anything estimates relative depth from one RGB image using Depth Anything V2. It gives you a depth map without a stereo pair or depth sensor.
Place Depth Anything under Analyze in the Image view. Select the Depth Anything V2 Small ONNX model and its matching runner configuration, then connect a picture to in.
With that configuration loaded, depth supplies the map at the input image resolution. Connect it to the field input of Depth Lens and adjust Focus against the map’s values.
Inputs and outputs
Section titled “Inputs and outputs”inis the source image.- The supplied Depth Anything configuration publishes
depth, a single-channel relative depth image.
Ports change when the configuration loads. The base declaration in the table below uses out.
Controls
Section titled “Controls”- ONNX Model selects the model file.
- Runner Config selects the matching configuration. Reload Config applies changes to it.
- The supplied configuration adds Normalize. On rescales each frame to 0 through 1 using that frame’s minimum and maximum. Off retains the raw relative values.
The map does not measure distance in metres. Per-frame normalization can shift the range when objects enter or leave the picture. Video Depth Anything is a separate operator that uses earlier frames to improve temporal consistency.
Inputs
-
inimage
Outputs
-
outimage
Parameters
Section titled “Parameters”Depth
op('depth_anything').par.Modelpath - Default:
"" (Empty String)
op('depth_anything').par.Configpath - Default:
"" (Empty String)
op('depth_anything').par.Reload - Default:
None