Adding text dataset generator]

import matplotlib.pyplot as plt

import numpy as np

from PIL import Image, ImageDraw, ImageFont

import random

# Number of points influence the resolution of your chosen text in the dataset

num_points = 10000

np.random.seed(42)

# Type chosen text

text = "NN"

# Add the custom text size

font_size = 600

# Change path to your chosen font

font_path = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"

######################################################################

# Define the coordinates of the rectangle's corners

x_min, y_min = -1, -1

x_max, y_max = 1, 1

# Create a figure and axis

fig, ax = plt.subplots()

# Remove the rectangle outline and background

ax.axis('off')

# Set explicit x-axis and y-axis limits

ax.set_xlim(x_min, x_max)

ax.set_ylim(y_min, y_max)

# Generate random points within the rectangle

random_x = np.random.uniform(x_min, x_max, num_points)

random_y = np.random.uniform(y_min, y_max, num_points)

# Create a blank image to render the "DNN" text

image_width = 1000

image_height = 1000

image = Image.new("L", (image_width, image_height), color=255)  # White background

draw = ImageDraw.Draw(image)

font = ImageFont.truetype(font_path, font_size)

text_width, text_height = draw.textsize(text, font=font)

text_x = (image_width - text_width) / 2

text_y = (image_height - text_height) / 2

draw.text((text_x, text_y), text, fill=0, font=font)  # Fill with black (0)

# Convert the image to a numpy array

text_image = np.array(image)

# Initialize labels list

labels = []

# Check if each random point is inside the text boundaries

for x, y in zip(random_x, random_y):

    # Convert x and y to image coordinates

    image_x = int((x - x_min) / (x_max - x_min) * image_width)

    image_y = int((y - y_min) / (y_max - y_min) * image_height)

    # Check if the pixel in the image is black (inside text)

    if text_image[image_y, image_x] == 0:

        labels.append(1)

    else:

        labels.append(-1)

# Filter the random_y array based on labels

inside_text_y = random_y[np.array(labels) == 1]

inside_text_y = inside_text_y[:1000]

# Filter the random_x array based on labels

inside_text_x = random_x[np.array(labels) == 1]

inside_text_x = inside_text_x[:1000]

# Filter the random_y array based on labels

outside_text_y = random_y[np.array(labels) == -1]

# Filter the random_x array based on labels

outside_text_x = random_x[np.array(labels) == -1]

# Plot the points inside the text

plt.scatter(inside_text_x, inside_text_y, c='g', marker='.', label='Class A')

plt.scatter(outside_text_x[:len(inside_text_x)], outside_text_y[:len(inside_text_y)], c='r', marker='.', label='Class B')

np.save("logo_dataset_classA", np.column_stack((inside_text_x, inside_text_y)))

np.save("logo_dataset_classB", np.column_stack((outside_text_x[:len(inside_text_x)], outside_text_y[:len(inside_text_y)])))

# Show the legend

#plt.legend()

# Show the plot

plt.show()