65 lines
2.3 KiB
Markdown
65 lines
2.3 KiB
Markdown
|
|
# Documentation
|
|
- Class name: DensePosePreprocessor
|
|
- Category: ControlNet Preprocessors/Faces and Poses Estimators
|
|
- Output node: False
|
|
|
|
DensePosePreprocessor节点旨在从图像中估算出密集的人体姿态。它使用选定的DensePose模型和颜色映射对图像进行预处理，以生成详细的身体姿态估计。
|
|
|
|
# Input types
|
|
## Required
|
|
- image
|
|
- 需要进行处理以估算密集姿态的输入图像。这是进行姿态估计的主要输入图像。
|
|
- Comfy dtype: IMAGE
|
|
- Python dtype: torch.Tensor
|
|
|
|
## Optional
|
|
- model
|
|
- 指定用于姿态估计的DensePose模型。模型的选择可能会影响姿态估计的准确性和性能。
|
|
- Comfy dtype: COMBO[STRING]
|
|
- Python dtype: List[str]
|
|
- cmap
|
|
- 决定用于可视化姿态估计结果的颜色映射。不同的颜色映射可以提供不同的视觉清晰度和美感。
|
|
- Comfy dtype: COMBO[STRING]
|
|
- Python dtype: List[str]
|
|
- resolution
|
|
- 处理前将输入图像调整到的分辨率。更高的分辨率可能会导致更详细的姿态估计，但可能会增加处理时间。
|
|
- Comfy dtype: INT
|
|
- Python dtype: int
|
|
|
|
# Output types
|
|
- image
|
|
- 输出一个根据所选颜色映射可视化的密集姿态估计图像。
|
|
- Comfy dtype: IMAGE
|
|
- Python dtype: Tuple[torch.Tensor]
|
|
|
|
|
|
## Usage tips
|
|
- Infra type: `GPU`
|
|
- Common nodes: unknown
|
|
|
|
|
|
## Source code
|
|
```python
|
|
class DensePose_Preprocessor:
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return create_node_input_types(
|
|
model=(["densepose_r50_fpn_dl.torchscript", "densepose_r101_fpn_dl.torchscript"], {"default": "densepose_r50_fpn_dl.torchscript"}),
|
|
cmap=(["Viridis (MagicAnimate)", "Parula (CivitAI)"], {"default": "Viridis (MagicAnimate)"})
|
|
)
|
|
|
|
RETURN_TYPES = ("IMAGE",)
|
|
FUNCTION = "execute"
|
|
|
|
CATEGORY = "ControlNet Preprocessors/Faces and Poses Estimators"
|
|
|
|
def execute(self, image, model, cmap, resolution=512):
|
|
from controlnet_aux.densepose import DenseposeDetector
|
|
model = DenseposeDetector \
|
|
.from_pretrained(filename=model) \
|
|
.to(model_management.get_torch_device())
|
|
return (common_annotator_call(model, image, cmap="viridis" if "Viridis" in cmap else "parula", resolution=resolution), )
|
|
|
|
```
|