86 lines
3.2 KiB
Markdown
86 lines
3.2 KiB
Markdown
|
|
# Documentation
|
|
- Class name: SaltLoadImageZip
|
|
- Category: SALT/Image/Loaders
|
|
- Output node: False
|
|
- Repo Ref: https://github.com/comfyanonymous/ComfyUI
|
|
|
|
SaltLoadImageZip节点专门用于从ZIP文件中加载并可选择性地调整图像大小。它支持多种图像格式，并可以将加载的图像连接成单个张量，便于在图像处理工作流程中使用。
|
|
|
|
# Input types
|
|
## Required
|
|
- path
|
|
- 指定包含图像的ZIP归档文件的文件路径。这对于定位和访问所需处理的图像至关重要。
|
|
- Comfy dtype: STRING
|
|
- Python dtype: str
|
|
- resize_images_to_first
|
|
- 决定是否将所有加载的图像调整为与找到的第一张图像尺寸相匹配。这确保了图像尺寸的一致性，这在批处理或将图像用于神经网络时通常是必需的。
|
|
- Comfy dtype: BOOLEAN
|
|
- Python dtype: bool
|
|
|
|
# Output types
|
|
- images
|
|
- 输出是一个包含已加载（可能已调整大小）图像的张量，可供进一步处理或分析使用。
|
|
- Comfy dtype: IMAGE
|
|
- Python dtype: torch.Tensor
|
|
|
|
|
|
## Usage tips
|
|
- Infra type: `GPU`
|
|
- Common nodes: unknown
|
|
|
|
|
|
## Source code
|
|
```python
|
|
class SaltLoadImageZip:
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"path": ("STRING", {}),
|
|
"resize_images_to_first": ("BOOLEAN", {"default": True})
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES = ("IMAGE",)
|
|
RETURN_NAMES = ("images",)
|
|
|
|
CATEGORY = "SALT/Image/Loaders"
|
|
FUNCTION = "load_images"
|
|
|
|
def load_images(self, path: str, resize_images_to_first: bool = True):
|
|
supported_formats = ('.png', '.jpg', '.jpeg', '.gif', '.tga', '.tiff', '.webp')
|
|
images = []
|
|
first_image_size = None
|
|
|
|
with zipfile.ZipFile(path, 'r') as z:
|
|
for file_name in z.namelist():
|
|
if file_name.lower().endswith(supported_formats):
|
|
with z.open(file_name) as file:
|
|
image = Image.open(BytesIO(file.read()))
|
|
if first_image_size is None:
|
|
first_image_size = image.size
|
|
if image.size == first_image_size or resize_images_to_first:
|
|
images.append(image if image.size == first_image_size else self.resize_right(image, first_image_size))
|
|
|
|
if not images:
|
|
raise ValueError(f"The input zip `{path}` does not contain any valid images!")
|
|
|
|
images = [pil2tensor(img) for img in images]
|
|
images = torch.cat(images, dim=0)
|
|
|
|
return (images, )
|
|
|
|
def resize_right(self, image, target_size):
|
|
img_ratio = image.width / image.height
|
|
target_ratio = target_size[0] / target_size[1]
|
|
resize_width, resize_height = (
|
|
(target_size[0], round(target_size[0] / img_ratio)) if target_ratio > img_ratio else
|
|
(round(target_size[1] * img_ratio), target_size[1])
|
|
)
|
|
image = image.resize((resize_width, resize_height), Image.Resampling.LANCZOS)
|
|
x_crop, y_crop = (resize_width - target_size[0]) // 2, (resize_height - target_size[1]) // 2
|
|
return image.crop((x_crop, y_crop, x_crop + target_size[0], y_crop + target_size[1]))
|
|
|
|
```
|