tst / image_processor.py
woodmastr's picture
Synced repo using 'sync_with_huggingface' Github Action
94cb743 verified
import argparse
import os
import shutil
from rembg import remove
from PIL import Image
import io
def add_background(image, background, default_color="#FFFFFF"):
"""
Adds a background to an image, with a fallback to a default color if the specified background is not available.
Args:
- image (PIL.Image.Image): Image with a transparent background.
- background (str or PIL.Image.Image): Background color (as a hex code) or a PIL Image to be used as background.
- default_color (str): Fallback color if the specified background is not valid. Defaults to white.
Returns:
- PIL.Image.Image: The image with the new background.
"""
foreground = image.convert("RGBA")
if isinstance(background, str) and (background.startswith("#") or background.isalpha()):
# Background is a color
try:
Image.new("RGBA", (1, 1), background) # Test if valid color
background_layer = Image.new("RGBA", foreground.size, background)
except ValueError:
print(
f"Invalid color '{background}'. Using default color '{default_color}'.")
background_layer = Image.new(
"RGBA", foreground.size, default_color)
elif isinstance(background, Image.Image):
# Background is an image
bg_img = background.convert("RGBA")
background_layer = bg_img.resize(foreground.size)
else:
# Fallback to default color
background_layer = Image.new("RGBA", foreground.size, default_color)
final_img = Image.alpha_composite(
background_layer, foreground).convert("RGB")
return final_img
def autocrop_image(image):
"""
Autocrops an image, focusing on the non-transparent pixels.
Args:
- image (PIL.Image.Image): Image to be autocropped.
Returns:
- PIL.Image.Image: The autocropped image.
"""
bbox = image.getbbox()
if bbox:
return image.crop(bbox)
return image
def remove_bg_func(image):
"""
Removes the background from an image using the rembg library.
Args:
- image (PIL.Image.Image): Image object from which to remove the background.
Returns:
- PIL.Image.Image: New image object with the background removed.
"""
# Convert the PIL Image to bytes
img_byte_arr = io.BytesIO()
image.save(img_byte_arr, format='PNG')
img_byte_arr = img_byte_arr.getvalue()
# Use rembg to remove the background
result_bytes = remove(img_byte_arr)
# Convert the result bytes back to a PIL Image
result_image = Image.open(io.BytesIO(result_bytes))
return result_image
def process_image(image_data, crop=False, remove_bg=False, resize=None, padding=0, background=None):
"""
Processes a single image based on the provided options.
Args:
- image_data (PIL.Image.Image): The input image.
- crop (bool): Whether to autocrop the image.
- remove_bg (bool): Whether to remove the background of the image.
- resize (tuple): Optional dimensions (width, height) to resize the image.
- padding (int): Number of padding pixels to add around the image.
- background (str): Optional background color (hex code or name) or path to an image file to set as the background.
Returns:
- PIL.Image.Image: The processed image.
"""
# Assume image_data is a PIL.Image.Image object
if remove_bg:
# Assuming remove_bg function returns a PIL image
image_data = remove_bg_func(image_data)
if crop:
# Assuming autocrop_image function modifies the image in place or returns a new PIL image
image_data = autocrop_image(image_data)
if resize:
# Assuming resize_and_pad_image function modifies the image in place or returns a new PIL image
image_data = resize_and_pad_image(image_data, resize, padding)
if background:
# Assuming add_background function modifies the image in place or returns a new PIL image
image_data = add_background(image_data, background)
return image_data
def resize_and_pad_image(image, dimensions, padding=0):
"""
Resizes an image to fit the specified dimensions and adds padding.
Args:
- image (PIL.Image.Image): Image object to be resized and padded.
- dimensions (tuple): Target dimensions (width, height).
- padding (int): Padding to add around the resized image.
Returns:
- PIL.Image.Image: Resized and padded image object.
"""
target_width, target_height = dimensions
content_width, content_height = target_width - \
2*padding, target_height - 2*padding
# Determine new size, preserving aspect ratio
img_ratio = image.width / image.height
target_ratio = content_width / content_height
if target_ratio > img_ratio:
new_height = content_height
new_width = int(new_height * img_ratio)
else:
new_width = content_width
new_height = int(new_width / img_ratio)
# Resize the image
resized_img = image.resize(
(new_width, new_height), Image.Resampling.LANCZOS)
# Create a new image with the target dimensions and a transparent background
new_img = Image.new(
"RGBA", (target_width, target_height), (255, 255, 255, 0))
# Calculate the position to paste the resized image to center it
paste_position = ((target_width - new_width) // 2,
(target_height - new_height) // 2)
# Paste the resized image onto the new image, centered
new_img.paste(resized_img, paste_position,
resized_img if resized_img.mode == 'RGBA' else None)
return new_img
def generate_output_filename(input_path, remove_bg=False, crop=False, resize=None, background=None):
"""
Generates an output filename based on the input path and processing options applied.
Appends specific suffixes based on the operations: '_b' for background removal, '_c' for crop,
and '_bg' if a background is added. It ensures the file extension is '.png'.
Args:
- input_path (str): Path to the input image.
- remove_bg (bool): Indicates if background removal was applied.
- crop (bool): Indicates if autocrop was applied.
- resize (tuple): Optional dimensions (width, height) for resizing the image.
- background (str): Indicates if a background was added (None if not used).
Returns:
- (str): Modified filename with appropriate suffix and '.png' extension.
"""
base, _ = os.path.splitext(os.path.basename(input_path))
suffix = ""
if remove_bg:
suffix += "_b"
if crop:
suffix += "_c"
if resize:
width, height = resize
suffix += f"_{width}x{height}"
if background:
suffix += "_bg" # Append "_bg" if the background option was used
# Ensure the file saves as PNG, accommodating for transparency or added backgrounds
return f"{base}{suffix}.png"
# The main and process_images functions remain the same, but ensure to update them to handle the new PNG output correctly.
# Update the process_images and main functions to include the new autocrop functionality
# Ensure to pass the crop argument to process_image and adjust the output filename generation accordingly
def process_images2(input_dir="./input", output_dir="./output", crop=False, remove_bg=False, resize=None, padding=0, background=None):
"""
Processes images in the specified directory based on the provided options.
Args:
- input_dir (str): Directory containing the images to be processed.
- output_dir (str): Directory where processed images will be saved.
- crop (bool): Whether to crop the images.
- remove_bg (bool): Whether to remove the background of the images.
- resize (tuple): Optional dimensions (width, height) to resize the image.
- padding (int): Number of padding pixels to add around the image.
- background (str): Optional background color (hex code or name) or path to an image file to set as the background.
"""
processed_input_dir = os.path.join(input_dir, "processed")
os.makedirs(processed_input_dir, exist_ok=True)
os.makedirs(output_dir, exist_ok=True)
inputs = [os.path.join(input_dir, f) for f in os.listdir(
input_dir) if os.path.isfile(os.path.join(input_dir, f))]
# if images are not in the input directory, print a message and return
if not inputs:
print("No images found in the input directory.")
return
for i, input_path in enumerate(inputs, start=1):
filename = os.path.basename(input_path)
output_filename = generate_output_filename(
input_path, remove_bg=remove_bg, crop=crop, resize=resize, background=background)
output_path = os.path.join(output_dir, output_filename)
print(f"Processing image {i}/{len(inputs)}...{filename}")
# Update the call to process_image with all parameters including background
process_image(input_path, output_path, crop=crop, remove_bg=remove_bg,
resize=resize, padding=padding, background=background)
shutil.move(input_path, os.path.join(processed_input_dir, filename))
print("All images have been processed.")
def process_images(input_dir="./input", output_dir="./output", crop=False, remove_bg=False, resize=None, padding=0, background=None):
"""
Processes images in the specified directory based on the provided options.
"""
processed_input_dir = os.path.join(input_dir, "processed")
os.makedirs(processed_input_dir, exist_ok=True)
os.makedirs(output_dir, exist_ok=True)
inputs = [os.path.join(input_dir, f) for f in os.listdir(
input_dir) if os.path.isfile(os.path.join(input_dir, f))]
if not inputs:
print("No images found in the input directory.")
return
for i, input_path in enumerate(inputs, start=1):
try:
with Image.open(input_path) as img:
# Define filename here, before it's used
filename = os.path.basename(input_path)
# Process the image
processed_img = process_image(
img, crop=crop, remove_bg=remove_bg, resize=resize, padding=padding, background=background)
# Generate output filename based on processing parameters
output_filename = generate_output_filename(
filename, remove_bg=remove_bg, crop=crop, resize=resize, background=background)
output_path = os.path.join(output_dir, output_filename)
# Save the processed image to the output directory
processed_img.save(output_path)
print(
f"Processed image {i}/{len(inputs)}: {filename} -> {output_filename}")
# Optionally move the processed input image to a "processed" subdirectory
shutil.move(input_path, os.path.join(
processed_input_dir, filename))
except Exception as e:
print(f"Error processing image {input_path}: {e}")
print("All images have been processed.")