BEAR-benchmark / util /concate_image.py
yqi19's picture
Add runnable eval code: API/local runners, GPT-judge scorer, per-task run.sh, util, README
ab9dacf verified
Raw
History Blame Contribute Delete
5.81 kB
from PIL import Image, ImageDraw, ImageFont
import numpy as np
def get_contrasting_color(image, x, y, width, height):
"""
Determine a contrasting color (black or white) based on the average color of a specified area in the image.
"""
# Crop the relevant part of the image
cropped_image = image.crop((x, y, x + width, y + height))
# Convert to numpy array for analysis
np_image = np.array(cropped_image)
# Calculate the average color
average_color = np.mean(np_image, axis=(0, 1))
# Brightness calculation based on perceived luminance
brightness = np.sqrt(0.299 * average_color[0] ** 2 + 0.587 * average_color[1] ** 2 + 0.114 * average_color[2] ** 2)
# Return white for dark backgrounds and black for light backgrounds
return 'white' if brightness < 128 else 'black'
# def concatenate_image(images, rows, columns, separator_width=10):
# # Ensure we have the exact number of images needed
# if len(images) != rows * columns:
# raise ValueError(f"Expected {rows * columns} images, but got {len(images)}.")
# # Calculate the max width and height of images to standardize sizes
# max_width = max(img.width for img in images)
# max_height = max(img.height for img in images)
# # Resize images to the max width and height
# resized_images = [img.resize((max_width, max_height), Image.Resampling.LANCZOS) for img in images]
# # Calculate the total width and height for the combined image
# total_width = max_width * columns + separator_width * (columns - 1)
# total_height = max_height * rows + separator_width * (rows - 1)
# combined_image = Image.new('RGB', (total_width, total_height), color='white')
# # Place images in the specified grid
# x_offset = 0
# y_offset = 0
# for i, img in enumerate(resized_images):
# combined_image.paste(img, (x_offset, y_offset))
# if (i + 1) % columns == 0: # Move to the next row after the last column
# x_offset = 0
# y_offset += img.height + separator_width
# else: # Move to the next column
# x_offset += img.width + separator_width
# # Add numbers to each image for identification
# draw = ImageDraw.Draw(combined_image)
# try:
# font_size = (max_width + max_height) // 2 // 12
# font = ImageFont.load_default(size=font_size)
# except IOError:
# font = ImageFont.truetype("arial", 20)
# x_offset = 0
# y_offset = 0
# for i, img in enumerate(resized_images):
# text = str(i + 1)
# text_x = x_offset + 10
# text_y = y_offset + 10
# text_width, text_height = font_size, font_size
# font_color = get_contrasting_color(combined_image, text_x, text_y, text_width, text_height)
# draw.text((text_x, text_y), text, fill=font_color, font=font)
# if (i + 1) % columns == 0:
# x_offset = 0
# y_offset += img.height + separator_width
# else:
# x_offset += img.width + separator_width
# return combined_image
def concatenate_image(images, rows, columns, separator_width=10):
# Calculate the max width and height of all provided images
max_width = max(img.width for img in images)
max_height = max(img.height for img in images)
# Resize all images to a uniform size
resized_images = [img.resize((max_width, max_height), Image.Resampling.LANCZOS) for img in images]
# Calculate number of images needed
total_images_needed = rows * columns
padding_needed = total_images_needed - len(resized_images)
# Pad with blank white images if needed
if padding_needed > 0:
blank_img = Image.new('RGB', (max_width, max_height), color='white')
resized_images.extend([blank_img] * padding_needed)
# Create the blank canvas for the grid
total_width = max_width * columns + separator_width * (columns - 1)
total_height = max_height * rows + separator_width * (rows - 1)
combined_image = Image.new('RGB', (total_width, total_height), color='white')
# Paste images into grid
x_offset = 0
y_offset = 0
for i, img in enumerate(resized_images):
combined_image.paste(img, (x_offset, y_offset))
if (i + 1) % columns == 0:
x_offset = 0
y_offset += max_height + separator_width
else:
x_offset += max_width + separator_width
# Draw labels
# draw = ImageDraw.Draw(combined_image)
# try:
# font_size = (max_width + max_height) // 2 // 12
# font = ImageFont.load_default(size=font_size)
# except:
# font = ImageFont.truetype("arial", 20)
# x_offset = 0
# y_offset = 0
# for i in range(len(resized_images)):
# text = str(i + 1)
# text_x = x_offset + 10
# text_y = y_offset + 10
# text_width, text_height = font_size, font_size
# font_color = get_contrasting_color(combined_image, text_x, text_y, text_width, text_height)
# draw.text((text_x, text_y), text, fill=font_color, font=font)
# if (i + 1) % columns == 0:
# x_offset = 0
# y_offset += max_height + separator_width
# else:
# x_offset += max_width + separator_width
# Draw labels
draw = ImageDraw.Draw(combined_image)
try:
font_size = (max_width + max_height) // 2 // 12
font = ImageFont.load_default(size=font_size)
except:
font = ImageFont.truetype("arial", 20)
x_offset = 0
y_offset = 0
original_count = len(images) # Only label original images
for i in range(original_count):
text = str(i + 1)
text_x = x_offset + 10
text_y = y_offset + 10
text_width, text_height = font_size, font_size
font_color = get_contrasting_color(combined_image, text_x, text_y, text_width, text_height)
draw.text((text_x, text_y), text, fill=font_color, font=font)
if (i + 1) % columns == 0:
x_offset = 0
y_offset += max_height + separator_width
else:
x_offset += max_width + separator_width
return combined_image
# from PIL import Image
# # Load or create 6 images
# images = [Image.open(f'01_pointing.png')]*7
# # Concatenate into a 2x3 grid
# result = concatenate_image(images, rows=2, columns=4)
# # Save or display the result
# result.save("output.jpg")