Depth Processor

Overview

The Depth Processor (AbProcDepth) is designed to generate a depth map from an input RGBA PNG image using a specified depth estimation model. The processor can cache the generated depth map, optionally invert it, and outputs the resulting image as a PNG.


Example Usage

- processor:
    processor: "depth"
    inputs:
      source: "input/input.png" # Virtual filepath of the input image in storage
    params:
      cache: "cache/depth.png" # Path where the generated depth map can be cached
      method: "ZoeD_N" # Depth estimation method; options include: ZoeD_N, ZoeD_K, ZoeD_NK
    outputs:
      target: "output/output.png" # Path where the final depth map image will be saved

Inputs

Key Required Description
source Yes Virtual filepath to the input image in the storage layer (e.g., "input/input.png").

Params

Key Type Required Default Description
cache str Yes (none) Virtual filepath where the depth map should be cached (e.g., "cache/depth.png").
method str Yes (none) Specifies the depth estimation model to use. Options include: "ZoeD_N", "ZoeD_K", and "ZoeD_NK".

Outputs

Key Required Description
target Virtual filepath where the processed depth map image will be saved.

Notes

  1. Model Selection:

    • The method parameter controls which depth estimation model is used.
    • Currently supported methods include "ZoeD_N", "ZoeD_K", and "ZoeD_NK". Additional methods may be added in the future.
  2. Caching:

    • The cache parameter allows for reuse on subsequent processing.
  3. Output Format:

    • The processor outputs the final depth map as a PNG in RGBA format, ensuring consistency with the rest of the image processing pipeline.