NN.py

import torch
from torch import nn, optim
from torchvision import datasets, transforms, models
from torch.utils.data import DataLoader, random_split
import sklearn
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix, classification_report
import numpy as np

import matplotlib.pyplot as plt


if __name__ == '__main__':  
# Pfad zu den Daten
    data_dir = 'data1'  # Passe den Pfad zu deinem Datensatz an

# Definition von Daten-Transformationen
    data_transform = transforms.Compose([
         transforms.RandomResizedCrop(224),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ])

    # Laden der Daten
    image_dataset = datasets.ImageFolder(root=data_dir, transform=data_transform)

    # Aufteilung der Daten in Trainings-, Validierungs- und Testdaten
    train_size = int(0.8 * len(image_dataset))
    val_size = test_size = (len(image_dataset) - train_size) // 2
    train_dataset, val_dataset, labels = random_split(image_dataset, [train_size, val_size, test_size])

    # Definition der DataLoaders
    train_loader = DataLoader(train_dataset, batch_size=4, shuffle=True, num_workers=4)
    val_loader = DataLoader(val_dataset, batch_size=4, shuffle=False, num_workers=4)
    test_loader = DataLoader(labels, batch_size=4, shuffle=False, num_workers=4)

    # Verwendung eines vortrainierten ResNet-Modells
    model = models.resnet18(weights=models.ResNet18_Weights.IMAGENET1K_V1)

    # Anpassung des Klassifikationslayers für die Anzahl der Klassen in deinem Datensatz
    num_classes = len(image_dataset.classes)
    model.fc = nn.Linear(model.fc.in_features, num_classes)

    # Definition der Verlustfunktion und des Optimierers
    criterion = nn.CrossEntropyLoss()
    optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)

    # Überprüfe, ob CUDA verfügbar ist, und weise das entsprechende Gerät zu
    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
    model.to(device)

    # Training des Modells
    num_epochs = 3  # Anpassen der Anzahl der Epochen nach Bedarf

    for epoch in range(num_epochs):
        for phase in ['train', 'val']:
            if phase == 'train':
                model.train()
                data_loader = train_loader
            else:
                model.eval()
                data_loader = val_loader if phase == 'val' else test_loader

            running_loss = 0.0
            running_corrects = 0

            for inputs, labels in data_loader:
                inputs, labels = inputs.to(device), labels.to(device)
                optimizer.zero_grad()

                with torch.set_grad_enabled(phase == 'train'):
                    outputs = model(inputs)
                    loss = criterion(outputs, labels)

                    if phase == 'train':
                        loss.backward()
                        optimizer.step()

                _, predictions = torch.max(outputs, 1)
                running_loss += loss.item() * inputs.size(0)
                running_corrects += torch.sum(predictions == labels.data)

            epoch_loss = running_loss / len(data_loader.dataset)
            epoch_acc = running_corrects.double() / len(data_loader.dataset)

            print(f'{phase} Loss: {epoch_loss:.4f} Acc: {epoch_acc:.4f}')


    print("Training abgeschlossen.")

    def get_predictions(model, data_loader, device):
        model.eval()
        all_predictions = []
        all_labels = []

        with torch.no_grad():
            for inputs, labels in data_loader:
                inputs = inputs.to(device)
                labels = labels.to(device)
                outputs = model(inputs)
                _, predictions = torch.max(outputs, 1)
                all_predictions.extend(predictions.cpu().numpy())
                all_labels.extend(labels.cpu().numpy())

        return np.array(all_predictions), np.array(all_labels)

    predictions, labels = get_predictions(model, test_loader, device)

    cm = sklearn.metrics.confusion_matrix(labels, predictions)

    ps = sklearn.metrics.precision_score(labels, predictions, average=None)

    f1 = sklearn.metrics.f1_score(labels, predictions, average=None)

    rc = sklearn.metrics.recall_score(labels, predictions, average=None)

    print('confusion matrix:', cm , "\n \n", 
        'precision:', ps ,"\n \n",
            'f1:', f1 , "\n \n",
            'recall:', rc , "\n \n")

    cmd = sklearn.metrics.ConfusionMatrixDisplay(cm, display_labels=["Apple_scab",
        "Apple_Black_rot",
               ])

    cmd.plot()
    plt.show()