Name:Muhammad Hermen Khan

Enroll:01-136221-054

BS-AI-6A

Dataset Link:https://www.kaggle.com/datasets/kaggleashwin/vehicle-type-recognition/data

!unzip /content/vehicles.zip -d /content/final

from sklearn.metrics import precision_recall_fscore_support,

accuracy_score
import numpy as np

from torch.utils.data import DataLoader, random_split

from torchvision import transforms, datasets
from torchvision import models
from torch import nn, optim
import torch

# Step 1: Dataset Preparation

# Dataset path
data_dir = r"/content/final/Dataset" # Update this to your dataset
path

# Define transformations
transform = transforms.Compose([
transforms.Resize((224, 224)), # Standard input size for most
models
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229,
0.224, 0.225])
])

# Load complete dataset

dataset = datasets.ImageFolder(data_dir, transform=transform)

# Split dataset into training (70%) and testing (30%)

train_size = int(0.7 * len(dataset))
test_size = len(dataset) - train_size
train_dataset, test_dataset = random_split(dataset, [train_size,
test_size])

# Create DataLoaders
batch_size = 32
train_loader = DataLoader(train_dataset, batch_size=batch_size,
shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=batch_size,
shuffle=False)

# Number of classes
num_classes = len(dataset.classes)
print(f"Number of classes: {num_classes}")
print(f"Classes: {dataset.classes}")

# Step 2: AlexNet Model

# Initialize the device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")

# Updated evaluation function to include precision, recall, F1-score

def train_and_evaluate_AlexNet(train_loader, test_loader, num_classes,
epochs=5):
# Load pretrained AlexNet model
model = models.alexnet(pretrained=True).to(device)

# Modify the classifier for the number of classes in the dataset

model.classifier[6] = nn.Linear(model.classifier[6].in_features,
num_classes).to(device)

# Define loss function and optimizer

criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

# Training loop
for epoch in range(epochs):
model.train()
running_loss = 0.0
for inputs, labels in train_loader:
inputs, labels = inputs.to(device), labels.to(device)
optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
running_loss += loss.item()
print(f"Epoch {epoch+1}/{epochs} - Loss:
{running_loss/len(train_loader):.4f}")

# Evaluation
model.eval()
all_preds = []
all_labels = []
with torch.no_grad():
for inputs, labels in test_loader:
inputs, labels = inputs.to(device), labels.to(device)
outputs = model(inputs)
_, preds = torch.max(outputs, 1)
all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())
 # Calculate metrics
all_preds = np.array(all_preds)
all_labels = np.array(all_labels)

accuracy = accuracy_score(all_labels, all_preds) * 100

precision, recall, f1, _ =
precision_recall_fscore_support(all_labels, all_preds,
average='weighted')

print(f"AlexNet Accuracy on Test Set: {accuracy:.2f}%")

print(f"Precision: {precision:.4f}")
print(f"Recall: {recall:.4f}")
print(f"F1-Score: {f1:.4f}")

# Call the function

train_and_evaluate_AlexNet(train_loader, test_loader, num_classes,
epochs=5)

Number of classes: 4
Classes: ['Bus', 'Car', 'Truck', 'motorcycle']
Using device: cuda

/usr/local/lib/python3.10/dist-packages/torchvision/models/
_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated
since 0.13 and may be removed in the future, please use 'weights'
instead.
warnings.warn(
/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:2
23: UserWarning: Arguments other than a weight enum or `None` for
'weights' are deprecated since 0.13 and may be removed in the future.
The current behavior is equivalent to passing
`weights=AlexNet_Weights.IMAGENET1K_V1`. You can also use
`weights=AlexNet_Weights.DEFAULT` to get the most up-to-date weights.
warnings.warn(msg)
Downloading: "https://download.pytorch.org/models/alexnet-owt-
7be5be79.pth" to /root/.cache/torch/hub/checkpoints/alexnet-owt-
7be5be79.pth
100%|██████████| 233M/233M [00:01<00:00, 150MB/s]
/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Epoch 1/5 - Loss: 1.6853

Epoch 2/5 - Loss: 1.3936
Epoch 3/5 - Loss: 1.3958
Epoch 4/5 - Loss: 1.3907
Epoch 5/5 - Loss: 1.3893
AlexNet Accuracy on Test Set: 22.50%
Precision: 0.0506
Recall: 0.2250
F1-Score: 0.0827

/usr/local/lib/python3.10/dist-packages/sklearn/metrics/
_classification.py:1531: UndefinedMetricWarning: Precision is ill-
defined and being set to 0.0 in labels with no predicted samples. Use
`zero_division` parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

from torchvision import models

from torch import nn, optim
import torch
from tqdm import tqdm
from sklearn.metrics import precision_recall_fscore_support,
accuracy_score
import numpy as np

# Define the device

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")

# Training and evaluation function for VGG16

def train_and_evaluate_VGG16(train_loader, test_loader, num_classes,
epochs=5):
# Load the pretrained VGG16 model
model = models.vgg16(pretrained=True).to(device)

# Modify the classifier to match the number of classes in the

dataset
model.classifier[6] = nn.Linear(model.classifier[6].in_features,
num_classes).to(device)

# Define the loss function and optimizer

criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

# Training loop
for epoch in range(epochs):
model.train()
running_loss = 0.0
for inputs, labels in tqdm(train_loader, desc=f"Epoch
{epoch+1}/{epochs}"):
inputs, labels = inputs.to(device), labels.to(device)
optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
 running_loss += loss.item()

print(f"Epoch {epoch+1}/{epochs} - Loss:

{running_loss/len(train_loader):.4f}")

# Evaluation
model.eval()
all_preds = []
all_labels = []
with torch.no_grad():
for inputs, labels in test_loader:
inputs, labels = inputs.to(device), labels.to(device)
outputs = model(inputs)
_, preds = torch.max(outputs, 1)
all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())

# Calculate metrics
all_preds = np.array(all_preds)
all_labels = np.array(all_labels)

accuracy = accuracy_score(all_labels, all_preds) * 100

precision, recall, f1, _ =
precision_recall_fscore_support(all_labels, all_preds,
average='weighted')

print(f"VGG16 Accuracy on Test Set: {accuracy:.2f}%")

print(f"Precision: {precision:.4f}")
print(f"Recall: {recall:.4f}")
print(f"F1-Score: {f1:.4f}")

# Call the function

# Ensure train_loader, test_loader, and num_classes are already
defined
train_and_evaluate_VGG16(train_loader, test_loader, num_classes,
epochs=5)

Using device: cuda

/usr/local/lib/python3.10/dist-packages/torchvision/models/
_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated
since 0.13 and may be removed in the future, please use 'weights'
instead.
warnings.warn(
/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:2
23: UserWarning: Arguments other than a weight enum or `None` for
'weights' are deprecated since 0.13 and may be removed in the future.
The current behavior is equivalent to passing
`weights=VGG16_Weights.IMAGENET1K_V1`. You can also use
`weights=VGG16_Weights.DEFAULT` to get the most up-to-date weights.
 warnings.warn(msg)
Downloading: "https://download.pytorch.org/models/vgg16-397923af.pth"
to /root/.cache/torch/hub/checkpoints/vgg16-397923af.pth
100%|██████████| 528M/528M [00:03<00:00, 157MB/s]
Epoch 1/5: 67%|██████▋ | 6/9 [00:10<00:05,
1.72s/it]/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
Epoch 1/5: 100%|██████████| 9/9 [00:14<00:00, 1.56s/it]

Epoch 1/5 - Loss: 3.5817

Epoch 2/5: 100%|██████████| 9/9 [00:13<00:00, 1.47s/it]

Epoch 2/5 - Loss: 1.4097

Epoch 3/5: 100%|██████████| 9/9 [00:12<00:00, 1.44s/it]

Epoch 3/5 - Loss: 1.4017

Epoch 4/5: 100%|██████████| 9/9 [00:13<00:00, 1.46s/it]

Epoch 4/5 - Loss: 1.4482

Epoch 5/5: 100%|██████████| 9/9 [00:13<00:00, 1.47s/it]

Epoch 5/5 - Loss: 1.3835

VGG16 Accuracy on Test Set: 26.67%
Precision: 0.1221
Recall: 0.2667
F1-Score: 0.1671

/usr/local/lib/python3.10/dist-packages/sklearn/metrics/
_classification.py:1531: UndefinedMetricWarning: Precision is ill-
defined and being set to 0.0 in labels with no predicted samples. Use
`zero_division` parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

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

# Define the LeNet model

class LeNet(nn.Module):
def __init__(self, num_classes):
 super(LeNet, self).__init__()
self.conv1 = nn.Conv2d(3, 6, kernel_size=5) # 3 input
channels (RGB)
self.conv2 = nn.Conv2d(6, 16, kernel_size=5)
self.fc1 = nn.Linear(16 * 53 * 53, 120) # Adjust input size
for your dataset
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, num_classes)

def forward(self, x):

x = torch.relu(self.conv1(x))
x = torch.max_pool2d(x, kernel_size=2, stride=2)
x = torch.relu(self.conv2(x))
x = torch.max_pool2d(x, kernel_size=2, stride=2)
x = x.view(x.size(0), -1) # Flatten for fully connected
layers
x = torch.relu(self.fc1(x))
x = torch.relu(self.fc2(x))
x = self.fc3(x)
return x

# Define the device

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")

# Define the training and evaluation function

def train_and_evaluate_LeNet(train_loader, test_loader, num_classes,
epochs=5):
# Initialize the LeNet model
model = LeNet(num_classes).to(device)

# Define the loss function and optimizer

criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

# Training loop
for epoch in range(epochs):
model.train()
running_loss = 0.0
for inputs, labels in tqdm(train_loader, desc=f"Epoch
{epoch+1}/{epochs}"):
inputs, labels = inputs.to(device), labels.to(device)
optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
running_loss += loss.item()
print(f"Epoch {epoch+1}/{epochs} - Loss:
{running_loss/len(train_loader):.4f}")
 # Evaluation
model.eval()
all_preds = []
all_labels = []
with torch.no_grad():
for inputs, labels in test_loader:
inputs, labels = inputs.to(device), labels.to(device)
outputs = model(inputs)
_, preds = torch.max(outputs, 1)
all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())

# Calculate metrics
all_preds = np.array(all_preds)
all_labels = np.array(all_labels)

accuracy = accuracy_score(all_labels, all_preds) * 100

precision, recall, f1, _ =
precision_recall_fscore_support(all_labels, all_preds,
average='weighted')

print(f"LeNet Accuracy on Test Set: {accuracy:.2f}%")

print(f"Precision: {precision:.4f}")
print(f"Recall: {recall:.4f}")
print(f"F1-Score: {f1:.4f}")

# Dataset Preparation
data_dir = r"/content/final/Dataset" # Update with your dataset path
transform = transforms.Compose([
transforms.Resize((224, 224)), # Resize for LeNet input
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229,
0.224, 0.225])
])

dataset = datasets.ImageFolder(data_dir, transform=transform)

# Split into training and testing sets

train_size = int(0.7 * len(dataset))
test_size = len(dataset) - train_size
train_dataset, test_dataset = random_split(dataset, [train_size,
test_size])

# Create DataLoaders
batch_size = 32
train_loader = DataLoader(train_dataset, batch_size=batch_size,
shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=batch_size,
shuffle=False)
# Number of classes
num_classes = len(dataset.classes)
print(f"Number of classes: {num_classes}")
print(f"Classes: {dataset.classes}")

# Train and evaluate LeNet

train_and_evaluate_LeNet(train_loader, test_loader, num_classes,
epochs=5)

Using device: cuda

Number of classes: 4
Classes: ['Bus', 'Car', 'Truck', 'motorcycle']

Epoch 1/5: 22%|██▏ | 2/9 [00:01<00:03,

1.75it/s]/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
Epoch 1/5: 100%|██████████| 9/9 [00:08<00:00, 1.06it/s]

Epoch 1/5 - Loss: 1.3808

Epoch 2/5: 100%|██████████| 9/9 [00:09<00:00, 1.04s/it]

Epoch 2/5 - Loss: 1.0704

Epoch 3/5: 100%|██████████| 9/9 [00:10<00:00, 1.16s/it]

Epoch 3/5 - Loss: 0.7165

Epoch 4/5: 100%|██████████| 9/9 [00:09<00:00, 1.04s/it]

Epoch 4/5 - Loss: 0.5021

Epoch 5/5: 100%|██████████| 9/9 [00:08<00:00, 1.11it/s]

Epoch 5/5 - Loss: 0.2552

LeNet Accuracy on Test Set: 54.17%
Precision: 0.5266
Recall: 0.5417
F1-Score: 0.5285

import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader, random_split
from torchvision import datasets, transforms
from torch.autograd import Variable
from tqdm import tqdm
import os
from PIL import Image
# 1. ZFNet Model Definition
class ZFNet(nn.Module):
def __init__(self, num_classes):
super(ZFNet, self).__init__()
# Convolutional Layers
self.conv1 = nn.Conv2d(3, 96, kernel_size=7, stride=2,
padding=1) # Output: 96 x 111 x 111
self.conv2 = nn.Conv2d(96, 256, kernel_size=5, stride=2,
padding=1) # Output: 256 x 27 x 27
self.conv3 = nn.Conv2d(256, 384, kernel_size=3, padding=1) #
Output: 384 x 13 x 13
self.conv4 = nn.Conv2d(384, 384, kernel_size=3, padding=1) #
Output: 384 x 13 x 13
self.conv5 = nn.Conv2d(384, 256, kernel_size=3, padding=1) #
Output: 256 x 6 x 6

# Pooling Layer
self.pool = nn.MaxPool2d(kernel_size=3, stride=2)

# Placeholder for flattened size

self._to_linear = None

# Fully Connected Layers

self.fc1 = None
self.fc2 = nn.Linear(4096, 4096)
self.fc3 = nn.Linear(4096, num_classes)

def forward(self, x):

# Pass through convolutional layers
x = torch.relu(self.conv1(x))
x = self.pool(x)
x = torch.relu(self.conv2(x))
x = self.pool(x)
x = torch.relu(self.conv3(x))
x = torch.relu(self.conv4(x))
x = torch.relu(self.conv5(x))
x = self.pool(x)

# Dynamically set the size of fc1

if self._to_linear is None:
self._to_linear = x.view(x.size(0), -1).size(1)
self.fc1 = nn.Linear(self._to_linear, 4096).to(x.device)
# Initialize fc1 dynamically

# Flatten the tensor

x = x.view(x.size(0), -1)

# Pass through fully connected layers

x = torch.relu(self.fc1(x))
 x = torch.relu(self.fc2(x))
x = self.fc3(x)
return x

# 2. Dataset Preparation with Random Split

data_transforms = transforms.Compose([
transforms.Resize((224, 224)), # Resize images to 224x224
transforms.RandomHorizontalFlip(), # Augmentation: Flip images
randomly
transforms.ToTensor(), # Convert images to tensor
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229,
0.224, 0.225]) # Normalize to ImageNet stats
])

# Path to your dataset folder (with subfolders like Bus, Car, etc.)
data_dir = '/content/final/Dataset' # Change this path to your
dataset directory

# Load dataset using ImageFolder

dataset = datasets.ImageFolder(data_dir, transform=data_transforms)

# Calculate the size for train/val split (80% train, 20% validation)
train_size = int(0.8 * len(dataset))
val_size = len(dataset) - train_size

# Split the dataset into train and validation sets

train_dataset, val_dataset = random_split(dataset, [train_size,
val_size])

# DataLoaders for batch processing

train_loader = DataLoader(train_dataset, batch_size=32, shuffle=True)
val_loader = DataLoader(val_dataset, batch_size=32, shuffle=False)

# 3. Initialize the Model, Loss, and Optimizer

num_classes = len(dataset.classes) # Number of classes (e.g., 4
classes: Bus, Car, Motorcycle, Truck)
model = ZFNet(num_classes=num_classes)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

model = model.to(device)

# Loss and Optimizer

criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

from sklearn.metrics import precision_recall_fscore_support,

accuracy_score
import numpy as np
# 4. Training the Model with Metrics Calculation
def train_model_with_metrics(model, criterion, optimizer,
num_epochs=25):
best_model_wts = model.state_dict()
best_acc = 0.0

for epoch in range(num_epochs):

print(f'Epoch {epoch}/{num_epochs - 1}')
print('-' * 10)

# Training phase
model.train()
running_loss = 0.0
running_corrects = 0

for inputs, labels in tqdm(train_loader):

inputs, labels = inputs.to(device), labels.to(device)

optimizer.zero_grad()

# Forward
outputs = model(inputs)
loss = criterion(outputs, labels)

# Backward
loss.backward()
optimizer.step()

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

epoch_loss = running_loss / len(train_loader.dataset)

epoch_acc = running_corrects.double() /
len(train_loader.dataset)

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

# Validation phase
model.eval()
running_corrects = 0
all_preds = []
all_labels = []

for inputs, labels in val_loader:

inputs, labels = inputs.to(device), labels.to(device)

with torch.no_grad():
 outputs = model(inputs)
_, preds = torch.max(outputs, 1)

# Store predictions and labels for metrics

all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())

running_corrects += torch.sum(preds == labels.data)

# Calculate metrics
all_preds = np.array(all_preds)
all_labels = np.array(all_labels)

val_acc = accuracy_score(all_labels, all_preds) * 100

precision, recall, f1, _ =
precision_recall_fscore_support(all_labels, all_preds,
average='weighted')

print(f'Validation Accuracy: {val_acc:.2f}%')

print(f'Precision: {precision:.4f}')
print(f'Recall: {recall:.4f}')
print(f'F1-Score: {f1:.4f}')

# Deep copy the model if it's the best so far

if val_acc > best_acc:
best_acc = val_acc
best_model_wts = model.state_dict()

print(f'Best Validation Accuracy: {best_acc:.2f}%')

model.load_state_dict(best_model_wts)
return model

# Train the model

model = train_model_with_metrics(model, criterion, optimizer,
num_epochs=25)

Epoch 0/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.07s/it]

Train Loss: 1.6090 Acc: 0.2406

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
 _warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 23.75%

Precision: 0.0963
Recall: 0.2375
F1-Score: 0.1300
Epoch 1/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.00it/s]

Train Loss: 1.3903 Acc: 0.2656

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 25.00%

Precision: 0.1450
Recall: 0.2500
F1-Score: 0.1197
Epoch 2/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.00s/it]

Train Loss: 1.3892 Acc: 0.2438

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 23.75%

Precision: 0.0564
Recall: 0.2375
F1-Score: 0.0912
Epoch 3/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.01it/s]

Train Loss: 1.3818 Acc: 0.2563

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 21.25%

Precision: 0.1092
Recall: 0.2125
F1-Score: 0.1378
Epoch 4/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.01it/s]

Train Loss: 1.4098 Acc: 0.2625

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 26.25%

Precision: 0.0689
Recall: 0.2625
F1-Score: 0.1092
Epoch 5/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.00s/it]

Train Loss: 1.3906 Acc: 0.2469

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 26.25%

Precision: 0.0689
Recall: 0.2625
F1-Score: 0.1092
Epoch 6/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.00it/s]

Train Loss: 1.3880 Acc: 0.2250

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 23.75%

Precision: 0.0609
Recall: 0.2375
F1-Score: 0.0969
Epoch 7/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.07s/it]

Train Loss: 1.3881 Acc: 0.2281

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 27.50%

Precision: 0.1543
Recall: 0.2750
F1-Score: 0.1490
Epoch 8/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.01s/it]

Train Loss: 1.3804 Acc: 0.2531

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 32.50%

Precision: 0.2684
Recall: 0.3250
F1-Score: 0.2655
Epoch 9/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.00it/s]

Train Loss: 1.3446 Acc: 0.2969

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 48.75%

Precision: 0.4777
Recall: 0.4875
F1-Score: 0.4274
Epoch 10/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.01it/s]

Train Loss: 1.2988 Acc: 0.3688

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 45.00%

Precision: 0.3597
Recall: 0.4500
F1-Score: 0.3878
Epoch 11/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.01s/it]

Train Loss: 1.2850 Acc: 0.3656

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
/usr/local/lib/python3.10/dist-packages/sklearn/metrics/_classificatio
n.py:1531: UndefinedMetricWarning: Precision is ill-defined and being
set to 0.0 in labels with no predicted samples. Use `zero_division`
parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Validation Accuracy: 42.50%

Precision: 0.3272
Recall: 0.4250
F1-Score: 0.3669
Epoch 12/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.00s/it]

Train Loss: 1.2114 Acc: 0.4625

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 51.25%

Precision: 0.5125
Recall: 0.5125
F1-Score: 0.4668
Epoch 13/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.00it/s]

Train Loss: 1.1532 Acc: 0.4969

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 52.50%

Precision: 0.5232
Recall: 0.5250
F1-Score: 0.4825
Epoch 14/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.00it/s]

Train Loss: 1.0977 Acc: 0.5156

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 47.50%

Precision: 0.5138
Recall: 0.4750
F1-Score: 0.4652
Epoch 15/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.00s/it]

Train Loss: 1.0509 Acc: 0.5344

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 47.50%

Precision: 0.4834
Recall: 0.4750
F1-Score: 0.4201
Epoch 16/24
----------
100%|██████████| 10/10 [00:09<00:00, 1.02it/s]

Train Loss: 1.0008 Acc: 0.6031

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 53.75%

Precision: 0.5349
Recall: 0.5375
F1-Score: 0.5203
Epoch 17/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.01it/s]

Train Loss: 0.9009 Acc: 0.6188

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 57.50%

Precision: 0.6244
Recall: 0.5750
F1-Score: 0.5744
Epoch 18/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.00it/s]

Train Loss: 0.8418 Acc: 0.6375

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 63.75%

Precision: 0.6349
Recall: 0.6375
F1-Score: 0.6345
Epoch 19/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.00it/s]

Train Loss: 0.8674 Acc: 0.6531

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 57.50%

Precision: 0.6378
Recall: 0.5750
F1-Score: 0.5597
Epoch 20/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.01it/s]

Train Loss: 0.7596 Acc: 0.7125

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 57.50%

Precision: 0.5731
Recall: 0.5750
F1-Score: 0.5650
Epoch 21/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.00it/s]

Train Loss: 0.6961 Acc: 0.7563

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 60.00%

Precision: 0.6005
Recall: 0.6000
F1-Score: 0.5906
Epoch 22/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.01it/s]

Train Loss: 0.6176 Acc: 0.7563

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(
Validation Accuracy: 58.75%
Precision: 0.6079
Recall: 0.5875
F1-Score: 0.5611
Epoch 23/24
----------

100%|██████████| 10/10 [00:10<00:00, 1.00s/it]

Train Loss: 0.6368 Acc: 0.7344

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 62.50%

Precision: 0.6434
Recall: 0.6250
F1-Score: 0.6148
Epoch 24/24
----------

100%|██████████| 10/10 [00:09<00:00, 1.01it/s]

Train Loss: 0.5416 Acc: 0.7938

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Validation Accuracy: 66.25%

Precision: 0.6651
Recall: 0.6625
F1-Score: 0.6517
Best Validation Accuracy: 66.25%

import torch
from torch import nn, optim
from torchvision import datasets, transforms, models
from torch.utils.data import DataLoader, random_split
from tqdm import tqdm

# Dataset path
data_dir = r"/content/final/Dataset" # Update with your dataset path

# Define transformations (ResNet requires specific normalization)

transform = transforms.Compose([
transforms.Resize((224, 224)), # Resize for ResNet input
transforms.ToTensor(),
 transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229,
0.224, 0.225])
])

# Load dataset
dataset = datasets.ImageFolder(data_dir, transform=transform)

# Split dataset into training (70%) and testing (30%)

train_size = int(0.7 * len(dataset))
test_size = len(dataset) - train_size
train_dataset, test_dataset = random_split(dataset, [train_size,
test_size])

# Create DataLoaders
batch_size = 32
train_loader = DataLoader(train_dataset, batch_size=batch_size,
shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=batch_size,
shuffle=False)

# Number of classes
num_classes = len(dataset.classes)
print(f"Number of classes: {num_classes}")
print(f"Classes: {dataset.classes}")

# Define the device (GPU or CPU)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")

# Load a pre-trained ResNet model

model = models.resnet50(pretrained=True).to(device) # You can also
try resnet18, resnet34, etc.

# Modify the fully connected layer to match the number of classes

model.fc = nn.Linear(model.fc.in_features, num_classes).to(device)

# Define loss function and optimizer

criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)
from sklearn.metrics import precision_recall_fscore_support,
accuracy_score
import numpy as np

# Training and evaluation loop with metrics

def train_and_evaluate_resnet_with_metrics(train_loader, test_loader,
model, criterion, optimizer, num_classes, epochs=5):
for epoch in range(epochs):
model.train()
running_loss = 0.0
for inputs, labels in tqdm(train_loader, desc=f"Epoch
{epoch+1}/{epochs}"):
inputs, labels = inputs.to(device), labels.to(device)
optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
running_loss += loss.item()

print(f"Epoch {epoch+1}/{epochs} - Loss:

{running_loss/len(train_loader):.4f}")

# Evaluation
model.eval()
all_preds = []
all_labels = []

with torch.no_grad():
for inputs, labels in test_loader:
inputs, labels = inputs.to(device), labels.to(device)
outputs = model(inputs)
_, preds = torch.max(outputs, 1)

# Store predictions and labels for metrics calculation

all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())

# Calculate metrics
all_preds = np.array(all_preds)
all_labels = np.array(all_labels)

accuracy = accuracy_score(all_labels, all_preds) * 100

precision, recall, f1, _ =
precision_recall_fscore_support(all_labels, all_preds,
average='weighted')

print(f"\nEvaluation Metrics on Test Set:")

print(f"Accuracy: {accuracy:.2f}%")
print(f"Precision: {precision:.4f}")
print(f"Recall: {recall:.4f}")
print(f"F1-Score: {f1:.4f}")

# Call the updated function to train and evaluate the model with
metrics
train_and_evaluate_resnet_with_metrics(train_loader, test_loader,
model, criterion, optimizer, num_classes, epochs=5)

Number of classes: 4
Classes: ['Bus', 'Car', 'Truck', 'motorcycle']
Using device: cuda
/usr/local/lib/python3.10/dist-packages/torchvision/models/
_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated
since 0.13 and may be removed in the future, please use 'weights'
instead.
warnings.warn(
/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:2
23: UserWarning: Arguments other than a weight enum or `None` for
'weights' are deprecated since 0.13 and may be removed in the future.
The current behavior is equivalent to passing
`weights=ResNet50_Weights.IMAGENET1K_V1`. You can also use
`weights=ResNet50_Weights.DEFAULT` to get the most up-to-date weights.
warnings.warn(msg)
Downloading: "https://download.pytorch.org/models/resnet50-
0676ba61.pth" to /root/.cache/torch/hub/checkpoints/resnet50-
0676ba61.pth
100%|██████████| 97.8M/97.8M [00:00<00:00, 160MB/s]
Epoch 1/5: 100%|██████████| 9/9 [00:10<00:00, 1.19s/it]

Epoch 1/5 - Loss: 0.7649

Epoch 2/5: 100%|██████████| 9/9 [00:10<00:00, 1.15s/it]

Epoch 2/5 - Loss: 0.5734

Epoch 3/5: 100%|██████████| 9/9 [00:11<00:00, 1.22s/it]

Epoch 3/5 - Loss: 0.3695

Epoch 4/5: 100%|██████████| 9/9 [00:10<00:00, 1.22s/it]

Epoch 4/5 - Loss: 0.1897

Epoch 5/5: 100%|██████████| 9/9 [00:10<00:00, 1.21s/it]

Epoch 5/5 - Loss: 0.0882

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Evaluation Metrics on Test Set:

Accuracy: 79.17%
Precision: 0.8028
Recall: 0.7917
F1-Score: 0.7929

import torch
from torch import nn, optim
from torchvision import datasets, transforms, models
from torch.utils.data import DataLoader, random_split
from tqdm import tqdm

# Dataset path
data_dir = r"/content/final/Dataset" # Update with your dataset path

# Define transformations (VGG requires specific normalization)

transform = transforms.Compose([
transforms.Resize((224, 224)), # Resize for VGG16 input
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229,
0.224, 0.225])
])

# Load dataset
dataset = datasets.ImageFolder(data_dir, transform=transform)

# Split dataset into training (70%) and testing (30%)

train_size = int(0.7 * len(dataset))
test_size = len(dataset) - train_size
train_dataset, test_dataset = random_split(dataset, [train_size,
test_size])

# Create DataLoaders
batch_size = 32
train_loader = DataLoader(train_dataset, batch_size=batch_size,
shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=batch_size,
shuffle=False)

# Number of classes
num_classes = len(dataset.classes)
print(f"Number of classes: {num_classes}")
print(f"Classes: {dataset.classes}")

# Define the device (GPU or CPU)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")

# Load a pre-trained VGG16 model

model = models.vgg16(pretrained=True).to(device)

# Modify the fully connected layers to match the number of classes

model.classifier[6] = nn.Linear(model.classifier[6].in_features,
num_classes).to(device)

# Define loss function and optimizer

criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

from sklearn.metrics import precision_recall_fscore_support,

accuracy_score
import numpy as np

# Training and evaluation loop with metrics

def train_and_evaluate_vgg16_with_metrics(train_loader, test_loader,
model, criterion, optimizer, num_classes, epochs=5):
for epoch in range(epochs):
model.train()
running_loss = 0.0
for inputs, labels in tqdm(train_loader, desc=f"Epoch
{epoch+1}/{epochs}"):
inputs, labels = inputs.to(device), labels.to(device)
optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
running_loss += loss.item()

print(f"Epoch {epoch+1}/{epochs} - Loss:

{running_loss/len(train_loader):.4f}")

# Evaluation
model.eval()
all_preds = []
all_labels = []

with torch.no_grad():
for inputs, labels in test_loader:
inputs, labels = inputs.to(device), labels.to(device)
outputs = model(inputs)
_, preds = torch.max(outputs, 1)

# Store predictions and labels for metrics calculation

all_preds.extend(preds.cpu().numpy())
all_labels.extend(labels.cpu().numpy())

# Calculate metrics
all_preds = np.array(all_preds)
all_labels = np.array(all_labels)

accuracy = accuracy_score(all_labels, all_preds) * 100

precision, recall, f1, _ =
precision_recall_fscore_support(all_labels, all_preds,
average='weighted')

print(f"\nEvaluation Metrics on Test Set:")

print(f"Accuracy: {accuracy:.2f}%")
print(f"Precision: {precision:.4f}")
print(f"Recall: {recall:.4f}")
 print(f"F1-Score: {f1:.4f}")

# Call the updated function to train and evaluate the model with
metrics
train_and_evaluate_vgg16_with_metrics(train_loader, test_loader,
model, criterion, optimizer, num_classes, epochs=5)

Number of classes: 4
Classes: ['Bus', 'Car', 'Truck', 'motorcycle']
Using device: cuda

/usr/local/lib/python3.10/dist-packages/torchvision/models/
_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated
since 0.13 and may be removed in the future, please use 'weights'
instead.
warnings.warn(
/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:2
23: UserWarning: Arguments other than a weight enum or `None` for
'weights' are deprecated since 0.13 and may be removed in the future.
The current behavior is equivalent to passing
`weights=VGG16_Weights.IMAGENET1K_V1`. You can also use
`weights=VGG16_Weights.DEFAULT` to get the most up-to-date weights.
warnings.warn(msg)
Epoch 1/5: 100%|██████████| 9/9 [00:13<00:00, 1.49s/it]

Epoch 1/5 - Loss: 1.9105

Epoch 2/5: 100%|██████████| 9/9 [00:13<00:00, 1.46s/it]

Epoch 2/5 - Loss: 1.9666

Epoch 3/5: 100%|██████████| 9/9 [00:12<00:00, 1.44s/it]

Epoch 3/5 - Loss: 1.4009

Epoch 4/5: 100%|██████████| 9/9 [00:12<00:00, 1.42s/it]

Epoch 4/5 - Loss: 1.3920

Epoch 5/5: 100%|██████████| 9/9 [00:13<00:00, 1.48s/it]

Epoch 5/5 - Loss: 1.3519

/usr/local/lib/python3.10/dist-packages/PIL/Image.py:1054:
UserWarning: Palette images with Transparency expressed in bytes
should be converted to RGBA images
warnings.warn(

Evaluation Metrics on Test Set:

Accuracy: 42.50%
Precision: 0.4012
Recall: 0.4250
F1-Score: 0.3700

/usr/local/lib/python3.10/dist-packages/sklearn/metrics/
_classification.py:1531: UndefinedMetricWarning: Precision is ill-
defined and being set to 0.0 in labels with no predicted samples. Use
`zero_division` parameter to control this behavior.
_warn_prf(average, modifier, f"{metric.capitalize()} is",
len(result))

Comparative Analysis of Model Performance

Model Accuracy (%) Precision Recall F1-Score
AlexNet 22.50 0.0506 0.2250 0.0827
VGG16 26.67 0.1221 0.2667 0.1671
LeNet 54.17 0.5266 0.5417 0.5285
ZFNet 66.25 0.6651 0.6625 0.6517
ResNet 79.17 0.8028 0.7917 0.7929