CavinHuang-comfyui-nodes-docs/docs/Inference_Core_Zoe-DepthMapPreprocessor.md

Documentation

• Class name: Inference_Core_Zoe-DepthMapPreprocessor
• Category: ControlNet Preprocessors/Normal and Depth Estimators
• Output node: False

Zoe深度图预处理器节点旨在使用ZoeDepth模型处理图像以估计深度图。它利用预训练模型生成深度估计，增强对图像中空间关系的理解，可用于各种应用场景。

Input types

Required

• image
  • 需要进行深度图估计的输入图像。该图像将由ZoeDepth模型处理以生成深度估计。
  • Comfy dtype: IMAGE
  • Python dtype: torch.Tensor

Optional

• resolution
  • resolution参数指定深度图的期望输出分辨率。它允许调整深度估计的细节级别。
  • Comfy dtype: INT
  • Python dtype: int

Output types

• image
  • 输出是输入图像的深度图，提供逐像素的深度估计。
  • Comfy dtype: IMAGE
  • Python dtype: torch.Tensor

Usage tips

• Infra type: GPU
• Common nodes: unknown

Source code

class Zoe_Depth_Map_Preprocessor:
    @classmethod
    def INPUT_TYPES(s):
        return create_node_input_types()

    RETURN_TYPES = ("IMAGE",)
    FUNCTION = "execute"

    CATEGORY = "ControlNet Preprocessors/Normal and Depth Estimators"

    def execute(self, image, resolution=512, **kwargs):
        from controlnet_aux.zoe import ZoeDetector

        model = ZoeDetector.from_pretrained().to(model_management.get_torch_device())
        out = common_annotator_call(model, image, resolution=resolution)
        del model
        return (out, )