plot_draw.py

import torch
import torchvision.models as models
import torchvision.transforms as transforms
from torchvision.datasets import ImageFolder
from torch.utils.data import DataLoader
from sklearn.manifold import TSNE
import matplotlib.pyplot as plt
import os
import numpy as np
from PIL import Image
import tqdm
from torch.utils.data import DataLoader, Dataset
import random
random_seed = 42
random.seed(random_seed)

# 1. Load ResNet50 model and prepare data
resnet50 = models.resnet50(pretrained=True)
resnet50.eval()  # Set the model to evaluation mode

# Remove the last fully connected layer to use the model as a feature extractor
feature_extractor = torch.nn.Sequential(*list(resnet50.children())[:-1])

from torchvision import transforms

preprocess = transforms.Compose(
    [
        transforms.Resize((256, 256)),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.5], [0.5]),
    ]
)


# Directories containing the two sets of images
image_folder_gan = 'generated_images/generated_images_gan'
image_folder_diff = "generated_images/diffusion_generated_image_256"
image_folder_orginal = 'data/mask_33k'
import random
# 2. Extract features
features = []
labels = []  # 0 for images from first set, 1 for images from second set
def extract_features(image_folder, image_names,label, sample_size=None):
    # image_names = [f for f in os.listdir(image_folder) if os.path.isfile(os.path.join(image_folder, f))]
    
    # # If sample_size is specified, randomly select sample_size images
    # if sample_size and len(image_names) > sample_size:
    #     random.shuffle(image_names)
    #     image_names = image_names[:sample_size]
    
    for image_name in tqdm.tqdm(image_names):
        image_path = os.path.join(image_folder, image_name)
        if os.path.isfile(image_path):
            # Load and preprocess image
            image = Image.open(image_path).convert('RGB')
            input_tensor = preprocess(image)
            input_tensor = input_tensor.unsqueeze(0)  # Create a mini-batch as expected by the model

            # Extract features
            output = feature_extractor(input_tensor)
            output = output.view(output.size(0), -1)  # Flatten the output
            features.append(output)
            labels.append(label)

# Extract features for the first set
# f=open("data//original_mask.txt","w")
all_files=os.listdir("data//mask_33k")
random.shuffle(all_files)
image_names3=all_files[:1000]
# f.write(all_files)
# assert(False)
# image_names3=open("data//original_mask.txt","r").read().split("\n")
with torch.no_grad():
    extract_features(image_folder_gan, [x for x in os.listdir(image_folder_gan)],label=0)
    extract_features(image_folder_diff, [x for x in os.listdir(image_folder_diff)],label=1)
    extract_features(image_folder_orginal, image_names3,label=2, sample_size=1000)  # Randomly select 1000 images

# Concatenate all features and labels
features = torch.cat(features)
labels = np.array(labels)

# 3. t-SNE Transformation
tsne = TSNE(n_components=2, random_state=0)
features_2d = tsne.fit_transform(features)

# 4. Plot the data and save the plot
plt.figure(figsize=(10, 8))
plt.scatter(features_2d[labels == 0, 0], features_2d[labels == 0, 1], alpha=0.5, label='Generated', marker='o', c='blue')
plt.scatter(features_2d[labels == 2, 0], features_2d[labels == 2, 1], alpha=0.5, label='Original', marker='x', c='red')
plt.legend()
plt.title('t-SNE visualization of skin cancer mask generation(GAN)')

# Save the plot
plt.savefig('results/gan_generated_image_256.png')


# 4. Plot the data and save the plot
plt.figure(figsize=(10, 8))
plt.scatter(features_2d[labels == 1, 0], features_2d[labels == 1, 1], alpha=0.5, label='Generated', marker='o', c='green')
plt.scatter(features_2d[labels == 2, 0], features_2d[labels == 2, 1], alpha=0.5, label='Original', marker='x', c='red')
plt.legend()
plt.title('t-SNE visualization of skin cancer mask generation(diffusion)')

# Save the plot
plt.savefig('results/diffusion_generated_image_256.png')