Skip to content

Fake Scribble Preprocessor Provider (SEGS)

Documentation

  • Class name: FakeScribblePreprocessor_Provider_for_SEGS __Inspire
  • Category: InspirePack/SEGS/ControlNet
  • Output node: False

This node provides a preprocessing step for SEGS (semantic segmentation models) by applying a fake scribble effect. It's designed to prepare images for further processing by enhancing or modifying their features to better suit the requirements of SEGS models.

Input types

Required

  • safe
    • Determines whether the preprocessing should be performed in a 'safe' mode, affecting the execution and results by potentially altering the processing intensity or methods used.
    • Comfy dtype: BOOLEAN
    • Python dtype: bool

Output types

  • segs_preprocessor
    • Comfy dtype: SEGS_PREPROCESSOR
    • Outputs a preprocessed object ready for use with SEGS models, specifically tailored with a fake scribble effect.
    • Python dtype: HED_Preprocessor_wrapper

Usage tips

  • Infra type: CPU
  • Common nodes: unknown

Source code

class FakeScribblePreprocessor_Provider_for_SEGS(HEDPreprocessor_Provider_for_SEGS):
    def doit(self, safe):
        obj = HED_Preprocessor_wrapper(safe, "FakeScribblePreprocessor")
        return (obj, )