61 lines
1.7 KiB
Python
61 lines
1.7 KiB
Python
import torch
|
|
import torch.nn.functional as F
|
|
from typing import Tuple
|
|
|
|
class NodoFaceCropping:
|
|
"""
|
|
A custom node to crop an image to a 2:3 aspect ratio
|
|
"""
|
|
CATEGORY = "Image Processing"
|
|
RETURN_TYPES = ("IMAGE",)
|
|
FUNCTION = "execute"
|
|
NODE_DISPLAY_NAME = "Nodogoro Cropper (2:3 Ratio)"
|
|
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"image": ("IMAGE",),
|
|
},
|
|
}
|
|
|
|
def execute(self, image: torch.Tensor) -> Tuple[torch.Tensor]:
|
|
try:
|
|
# Print input shape for debugging
|
|
print(f"Original input shape: {image.shape}")
|
|
|
|
# Get dimensions
|
|
batch, height, width, channels = image.shape
|
|
|
|
# Calculate crop width for 2:3 ratio (keeping height constant)
|
|
target_width = int(height * (2/3))
|
|
|
|
# Calculate crop boundaries (center crop)
|
|
start_x = (width - target_width) // 2
|
|
end_x = start_x + target_width
|
|
|
|
# Perform the crop
|
|
cropped = image[:, :, start_x:end_x, :]
|
|
|
|
print(f"Final output shape: {cropped.shape}")
|
|
|
|
# Verify we have the correct ratio
|
|
assert cropped.shape[2] == int(cropped.shape[1] * (2/3)), "Output is not 2:3 ratio"
|
|
|
|
return (cropped,)
|
|
|
|
except Exception as e:
|
|
print(f"Error in cropping: {str(e)}")
|
|
print(f"Input shape: {image.shape}")
|
|
return (image,)
|
|
|
|
# Node class mappings
|
|
NODE_CLASS_MAPPINGS = {
|
|
"NodoFaceCropping": NodoFaceCropping
|
|
}
|
|
|
|
# Node display name mappings
|
|
NODE_DISPLAY_NAME_MAPPINGS = {
|
|
"NodoFaceCropping": "Face Cropper (2:3 Ratio)"
|
|
}
|