vit_model_training.py

import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
from transformers import ViTForImageClassification
from PIL import Image
import torch.optim as optim
import os
import pandas as pd
from sklearn.model_selection import train_test_split
## working 18.9.24


def labeling(path_real, path_fake):
    image_paths = []
    labels = []

    for filename in os.listdir(path_real):
        image_paths.append(os.path.join(path_real, filename))
        labels.append(0)

    for filename in os.listdir(path_fake):
        image_paths.append(os.path.join(path_fake, filename))
        labels.append(1)

    dataset = pd.DataFrame({'image_path': image_paths, 'label': labels})

    return dataset

class CustomDataset(Dataset):
    def __init__(self, dataframe, transform=None):
        self.dataframe = dataframe
        self.transform = transform

    def __len__(self):
        return len(self.dataframe)

    def __getitem__(self, idx):
        image_path = self.dataframe.iloc[idx, 0]  
        image = Image.open(image_path).convert('RGB')  # Convert to RGB format

        if self.transform:
            image = self.transform(image)

        label = self.dataframe.iloc[idx, 1]  
        return image, label
    
    
    
    
def shuffle_and_split_data(dataframe, test_size=0.2, random_state=59):
    # Shuffle the DataFrame
    shuffled_df = dataframe.sample(frac=1, random_state=random_state).reset_index(drop=True)
    
    # Split the DataFrame into train and validation sets
    train_df, val_df = train_test_split(shuffled_df, test_size=test_size, random_state=random_state)
    
    return train_df, val_df    


if __name__ == "__main__":
    # Check for GPU availability
    device = torch.device('cuda')

    # Load the pre-trained ViT model
    model = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224').to(device)

    # Freeze pre-trained layers
    for param in model.parameters():
        param.requires_grad = False

    # Define a new classifier
    model.classifier = nn.Linear(model.config.hidden_size, 2).to(device)  # Two output classes: 'REAL' and 'FAKE'
    
    print(model)
    # Define the optimizer
    optimizer = optim.Adam(model.parameters(), lr=0.001)

    # Resize image and make it a tensor (add dimension)
    preprocess = transforms.Compose([
        transforms.Resize((224, 224)),
        transforms.ToTensor()
    ])

    # Assuming you have already defined your dataset class and split it into training and validation sets
    # Let's call it CustomDataset



    train_real_folder = 'datasets/training_set/real'
    train_fake_folder = 'datasets/training_set/fake'





    train_dataset_df = labeling(train_real_folder, train_fake_folder)

    train_dataset_df , val_dataset_df =  shuffle_and_split_data(train_dataset_df)




    # Define the dataset and dataloaders
    train_dataset = CustomDataset(train_dataset_df, transform=preprocess)
    train_loader = DataLoader(train_dataset, batch_size=32, shuffle=True)

    val_dataset = CustomDataset(val_dataset_df, transform=preprocess)
    val_loader = DataLoader(val_dataset, batch_size=32)

    # Define the loss function and move it to GPU
    criterion = nn.CrossEntropyLoss().to(device)

    # Training loop
    num_epochs = 10
    for epoch in range(num_epochs):
        model.train()
        running_loss = 0.0
        for images, labels in train_loader:
            # Move inputs and labels to GPU
            images, labels = images.to(device), labels.to(device)
            
            optimizer.zero_grad()
            outputs = model(images)
            logits = outputs.logits  # Extract logits from the output
            loss = criterion(logits, labels)
            loss.backward()
            optimizer.step()
            running_loss += loss.item()
        print(f"Epoch {epoch+1}/{num_epochs}, Loss: {running_loss / len(train_loader)}")

        # Validation loop
        model.eval()
        correct = 0
        total = 0
        with torch.no_grad():
            for images, labels in val_loader:
                images, labels = images.to(device), labels.to(device)  # Move inputs and labels to GPU
                outputs = model(images)
                logits = outputs.logits  # Extract logits from the output
                _, predicted = torch.max(logits, 1)
                total += labels.size(0)
                correct += (predicted == labels).sum().item()
        print(f"Validation Accuracy: {correct / total}")

    # Save the trained model
    torch.save(model.state_dict(), 'trained_model.pth')