Project Mini

Project_Mini
March 6, 2024
0.1 Mobilenet
0.1.1 1. Load the Data
[2]: import torch
import torchvision.transforms as transforms
import torchvision.datasets as datasets
from torch.utils.data import DataLoader
# Define data transformations

transform = transforms.Compose([
transforms.Resize((224, 224)), # Resize to fit model input size
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])␣
↪ # ImageNet normalization
])
# Load dataset
train_dataset = datasets.ImageFolder(root='train', transform=transform)
test_dataset = datasets.ImageFolder(root='test', transform=transform)
# Create data loaders

train_loader = DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)
test_loader = DataLoader(dataset=test_dataset, batch_size=64, shuffle=False)
[15]: train_dataset.classes
[15]: ['001.Black_footed_Albatross',
'002.Laysan_Albatross',
'003.Sooty_Albatross',
'004.Groove_billed_Ani',
'005.Crested_Auklet',
'006.Least_Auklet',
'007.Parakeet_Auklet',
'008.Rhinoceros_Auklet',
'009.Brewer_Blackbird',
'010.Red_winged_Blackbird',
'011.Rusty_Blackbird',
'012.Yellow_headed_Blackbird',
1
'013.Bobolink',
'014.Indigo_Bunting',
'015.Lazuli_Bunting',
'016.Painted_Bunting',
'017.Cardinal',
'018.Spotted_Catbird',
'019.Gray_Catbird',
'020.Yellow_breasted_Chat',
'021.Eastern_Towhee',
'022.Chuck_will_Widow',
'023.Brandt_Cormorant',
'024.Red_faced_Cormorant',
'025.Pelagic_Cormorant',
'026.Bronzed_Cowbird',
'027.Shiny_Cowbird',
'028.Brown_Creeper',
'029.American_Crow',
'030.Fish_Crow',
'031.Black_billed_Cuckoo',
'032.Mangrove_Cuckoo',
'033.Yellow_billed_Cuckoo',
'034.Gray_crowned_Rosy_Finch',
'035.Purple_Finch',
'036.Northern_Flicker',
'037.Acadian_Flycatcher',
'038.Great_Crested_Flycatcher',
'039.Least_Flycatcher',
'040.Olive_sided_Flycatcher',
'041.Scissor_tailed_Flycatcher',
'042.Vermilion_Flycatcher',
'043.Yellow_bellied_Flycatcher',
'044.Frigatebird',
'045.Northern_Fulmar',
'046.Gadwall',
'047.American_Goldfinch',
'048.European_Goldfinch',
'049.Boat_tailed_Grackle',
'050.Eared_Grebe',
'051.Horned_Grebe',
'052.Pied_billed_Grebe',
'053.Western_Grebe',
'054.Blue_Grosbeak',
'055.Evening_Grosbeak',
'056.Pine_Grosbeak',
'057.Rose_breasted_Grosbeak',
'058.Pigeon_Guillemot',
'059.California_Gull',
2
'060.Glaucous_winged_Gull',
'061.Heermann_Gull',
'062.Herring_Gull',
'063.Ivory_Gull',
'064.Ring_billed_Gull',
'065.Slaty_backed_Gull',
'066.Western_Gull',
'067.Anna_Hummingbird',
'068.Ruby_throated_Hummingbird',
'069.Rufous_Hummingbird',
'070.Green_Violetear',
'071.Long_tailed_Jaeger',
'072.Pomarine_Jaeger',
'073.Blue_Jay',
'074.Florida_Jay',
'075.Green_Jay',
'076.Dark_eyed_Junco',
'077.Tropical_Kingbird',
'078.Gray_Kingbird',
'079.Belted_Kingfisher',
'080.Green_Kingfisher',
'081.Pied_Kingfisher',
'082.Ringed_Kingfisher',
'083.White_breasted_Kingfisher',
'084.Red_legged_Kittiwake',
'085.Horned_Lark',
'086.Pacific_Loon',
'087.Mallard',
'088.Western_Meadowlark',
'089.Hooded_Merganser',
'090.Red_breasted_Merganser',
'091.Mockingbird',
'092.Nighthawk',
'093.Clark_Nutcracker',
'094.White_breasted_Nuthatch',
'095.Baltimore_Oriole',
'096.Hooded_Oriole',
'097.Orchard_Oriole',
'098.Scott_Oriole',
'099.Ovenbird',
'100.Brown_Pelican',
'101.White_Pelican',
'102.Western_Wood_Pewee',
'103.Sayornis',
'104.American_Pipit',
'105.Whip_poor_Will',
'106.Horned_Puffin',
3
'107.Common_Raven',
'108.White_necked_Raven',
'109.American_Redstart',
'110.Geococcyx',
'111.Loggerhead_Shrike',
'112.Great_Grey_Shrike',
'113.Baird_Sparrow',
'114.Black_throated_Sparrow',
'115.Brewer_Sparrow',
'116.Chipping_Sparrow',
'117.Clay_colored_Sparrow',
'118.House_Sparrow',
'119.Field_Sparrow',
'120.Fox_Sparrow',
'121.Grasshopper_Sparrow',
'122.Harris_Sparrow',
'123.Henslow_Sparrow',
'124.Le_Conte_Sparrow',
'125.Lincoln_Sparrow',
'126.Nelson_Sharp_tailed_Sparrow',
'127.Savannah_Sparrow',
'128.Seaside_Sparrow',
'129.Song_Sparrow',
'130.Tree_Sparrow',
'131.Vesper_Sparrow',
'132.White_crowned_Sparrow',
'133.White_throated_Sparrow',
'134.Cape_Glossy_Starling',
'135.Bank_Swallow',
'136.Barn_Swallow',
'137.Cliff_Swallow',
'138.Tree_Swallow',
'139.Scarlet_Tanager',
'140.Summer_Tanager',
'141.Artic_Tern',
'142.Black_Tern',
'143.Caspian_Tern',
'144.Common_Tern',
'145.Elegant_Tern',
'146.Forsters_Tern',
'147.Least_Tern',
'148.Green_tailed_Towhee',
'149.Brown_Thrasher',
'150.Sage_Thrasher',
'151.Black_capped_Vireo',
'152.Blue_headed_Vireo',
'153.Philadelphia_Vireo',
4
'154.Red_eyed_Vireo',
'155.Warbling_Vireo',
'156.White_eyed_Vireo',
'157.Yellow_throated_Vireo',
'158.Bay_breasted_Warbler',
'159.Black_and_white_Warbler',
'160.Black_throated_Blue_Warbler',
'161.Blue_winged_Warbler',
'162.Canada_Warbler',
'163.Cape_May_Warbler',
'164.Cerulean_Warbler',
'165.Chestnut_sided_Warbler',
'166.Golden_winged_Warbler',
'167.Hooded_Warbler',
'168.Kentucky_Warbler',
'169.Magnolia_Warbler',
'170.Mourning_Warbler',
'171.Myrtle_Warbler',
'172.Nashville_Warbler',
'173.Orange_crowned_Warbler',
'174.Palm_Warbler',
'175.Pine_Warbler',
'176.Prairie_Warbler',
'177.Prothonotary_Warbler',
'178.Swainson_Warbler',
'179.Tennessee_Warbler',
'180.Wilson_Warbler',
'181.Worm_eating_Warbler',
'182.Yellow_Warbler',
'183.Northern_Waterthrush',
'184.Louisiana_Waterthrush',
'185.Bohemian_Waxwing',
'186.Cedar_Waxwing',
'187.American_Three_toed_Woodpecker',
'188.Pileated_Woodpecker',
'189.Red_bellied_Woodpecker',
'190.Red_cockaded_Woodpecker',
'191.Red_headed_Woodpecker',
'192.Downy_Woodpecker',
'193.Bewick_Wren',
'194.Cactus_Wren',
'195.Carolina_Wren',
'196.House_Wren',
'197.Marsh_Wren',
'198.Rock_Wren',
'199.Winter_Wren',
'200.Common_Yellowthroat']
5
0.1.2 2. Initialise the Training Model
Here, we are using the Mobile Net Architecture for the same since it has proved to be fast and at
the same time has good accuracy
[5]: import torch.nn as nn

import torchvision.models as models
# Initialize MobileNet model

mobilenet = models.mobilenet_v2(pretrained=True)
# Modify the last fully connected layer for the number of classes in your␣
↪dataset
num_classes = len(train_dataset.classes)
mobilenet.classifier[1] = nn.Linear(in_features=1280, out_features=num_classes)
# Freeze convolutional layers

for param in mobilenet.parameters():
param.requires_grad = False
# Unfreeze the last few layers for fine-tuning

for param in mobilenet.classifier.parameters():
param.requires_grad = True
C:\Users\ASUS\anaconda3\lib\site-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(
C:\Users\ASUS\anaconda3\lib\site-packages\torchvision\models\_utils.py:223:
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=MobileNet_V2_Weights.IMAGENET1K_V1`. You can also
use `weights=MobileNet_V2_Weights.DEFAULT` to get the most up-to-date weights.
warnings.warn(msg)
Downloading: "https://download.pytorch.org/models/mobilenet_v2-b0353104.pth" to
C:\Users\ASUS/.cache\torch\hub\checkpoints\mobilenet_v2-b0353104.pth
100%|��
��| 13.6M/13.6M [00:02<00:00, 5.51MB/s]
0.1.3 3. Define the Training Loop

[10]: import torch.optim as optim
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

mobilenet.to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(mobilenet.parameters(), lr=0.001)
6
def train_m(model, train_loader, criterion, optimizer, num_epochs=30):
train_losses = []
train_accuracy = []
for epoch in range(num_epochs):

model.train()
running_loss = 0.0
correct = 0
total = 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()
_, predicted = outputs.max(1)
total += labels.size(0)
correct += predicted.eq(labels).sum().item()
epoch_loss = running_loss / len(train_loader)

epoch_accuracy = 100. * correct / total
train_losses.append(epoch_loss)
train_accuracy.append(epoch_accuracy)
print(f"Epoch [{epoch+1}/{num_epochs}], Loss: {epoch_loss:.4f},␣

↪Accuracy: {epoch_accuracy:.2f}%")
return train_losses, train_accuracy
train_losses_m, train_accuracy_m = train_m(mobilenet, train_loader, criterion,␣

↪optimizer)
Epoch [1/30], Loss: 0.5475, Accuracy: 88.81%

7
0.1.4 4. Plot the training results
[11]: import matplotlib.pyplot as plt
# Plot architecture
print(mobilenet)
# Plot training details

plt.plot(train_losses_m, label='Train Loss')
plt.plot(train_accuracy_m, label='Train Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Value')
plt.title('Training Loss and Accuracy')
plt.legend()
plt.show()
MobileNetV2(
(features): Sequential(
(0): Conv2dNormActivation(
(0): Conv2d(3, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1),
bias=False)
(1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True,
track_running_stats=True)
(2): ReLU6(inplace=True)
)
8
(1): InvertedResidual(
(conv): Sequential(
groups=32, bias=False)
)
(1): Conv2d(32, 16, kernel_size=(1, 1), stride=(1, 1), bias=False)
)
)
(conv): Sequential(
)
groups=96, bias=False)
)
)
)
(conv): Sequential(
)
(0): Conv2d(144, 144, kernel_size=(3, 3), stride=(1, 1), padding=(1,
1), groups=144, bias=False)
9
)
)
)
(conv): Sequential(
)
)
)
)
(conv): Sequential(
)
)
)
)
(conv): Sequential(
10
)
)
)
)
(conv): Sequential(
)
)
)
)
(conv): Sequential(
)
11
)
)
)
(conv): Sequential(
)
)
)
)
(conv): Sequential(
)
)
12
)
)
(conv): Sequential(
)
)
)
)
(conv): Sequential(
)
)
)
)
(conv): Sequential(
13
)
)
)
)
(conv): Sequential(
)
)
)
)
(conv): Sequential(
)
14
)
)
)
(conv): Sequential(
)
)
)
)
(conv): Sequential(
)
)
)
)
15
)
)
(classifier): Sequential(
(0): Dropout(p=0.2, inplace=False)
(1): Linear(in_features=1280, out_features=200, bias=True)
)
)
[48]: device
[48]: device(type='cuda')
[12]: torch.save(mobilenet.state_dict(), 'mobilenet_checkpoint.pth')
0.1.5 5. Evaluate the Model

[13]: def evaluate_m(model, test_loader):
model.eval()
correct = 0
total = 0
16
with torch.no_grad():
for inputs, labels in test_loader:
_, predicted = torch.max(outputs, 1)
correct += (predicted == labels).sum().item()
accuracy = 100. * correct / total

print(f"Test Accuracy: {accuracy:.2f}%")
evaluate_m(mobilenet, test_loader)
Test Accuracy: 55.25%
0.2 Inception
[ ]:
0.2.1 1. Initialise the Training Model

Here, we are using the Inception Architecture for the same since it has proved to be reliable with
1*1 convolutions
[49]: import torch

# Load pre-trained Inception model

inception_model = models.inception_v3(pretrained=True)
inception_model.to(device)
# Modify the last fully connected layer for the desired number of classes
inception_model.fc = nn.Linear(inception_model.fc.in_features, num_classes)
# Freeze convolutional layers

for param in inception_model.parameters():
param.requires_grad = False
# Unfreeze the last few layers for fine-tuning

for param in inception_model.fc.parameters():
param.requires_grad = True
17
0.2.2 2. Load the Data
[53]: import torchvision.transforms as transforms
from torchvision.datasets import ImageFolder
# Define transformations for data preprocessing

transforms.Resize((299, 299)), # Resize images to match Inception input␣
↪size
transforms.ToTensor(), # Convert images to tensors

transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])␣
↪ # Normalize images
])
# Load training and testing datasets

train_dataset = ImageFolder(root='train/', transform=transform)
test_dataset = ImageFolder(root='test/', transform=transform)

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

[58]: import torch.optim as optim
import torch.nn as nn

inception_model.to(device)
optimizer = optim.Adam(inception_model.parameters(), lr=0.001)
def train(model, train_loader, criterion, optimizer, num_epochs=25):

train_losses = []
train_accuracy = []

model.train()
running_loss = 0.0
correct = 0
total = 0

18
# Resize input images to match Inception model's input size

inputs = nn.functional.interpolate(inputs, size=(299, 299),␣
↪mode='bilinear', align_corners=False)
# Forward pass
outputs, aux_outputs = model(inputs)
# Get main output (logits) from the 'output' tensor

main_outputs = outputs
# Calculate loss
loss = criterion(main_outputs, labels)
# Backward pass and optimization

loss.backward()
optimizer.step()
# Calculate accuracy
_, predicted = main_outputs.max(1)
correct += predicted.eq(labels).sum().item()


↪Accuracy: {epoch_accuracy:.2f}%")
[59]: train(inception_model, train_loader, criterion, optimizer, num_epochs=20)

19
[59]: ([4.510051303721489,
2.9086139341618154,
2.240238940462153,
1.8574077417241766,
1.634935606667336,
1.4252428710460663,
1.2798798058895355,
1.1566238967662161,
1.0691858834408698,
1.000948575265864,
0.947598285973072,
0.8681064024567604,
0.8172275359009175,
0.7825873259534227,
0.7598109075680692,
0.694906362669265,
0.702841191691287,
0.6455970829788674,
0.6202751841950924,
0.6106431853580983],
[12.178845512178846,
36.97030363697031,
49.1991991991992,
57.741074407741074,
61.94527861194528,
66.59993326659993,
69.70303636970304,
72.82282282282283,
74.02402402402403,
75.95929262595929,
76.24290957624291,
78.44511177844511,
79.12912912912913,
80.24691358024691,
80.53053053053053,
82.19886553219887,
20
81.2645979312646,
82.34901568234902,
83.60026693360027,
83.88388388388388])
[61]: torch.save(inception_model.state_dict(), 'inception_model.pth')

[63]: def evaluate(model, test_loader):
model.eval()
correct = 0
total = 0
accuracy = 100.*correct / total

print(f"Test Accuracy: {accuracy:.4f}")
evaluate(inception_model, test_loader)
Test Accuracy: 51.8813

[122]: import matplotlib.pyplot as plt
# Plot training loss and accuracy curves

plt.figure(figsize=(10, 5))
plt.subplot(1, 2, 1)
plt.plot(train_losses, label='Train Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title('Training Loss Curve')
plt.legend()
plt.plot(train_accuracy, label='Train Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy (%)')
plt.title('Training Accuracy Curve')
21
plt.legend()
plt.show()
0.3 EﬀicientNet
[ ]:
[66]: !pip install efficientnet_pytorch
Collecting efficientnet_pytorch
Downloading efficientnet_pytorch-0.7.1.tar.gz (21 kB)
Preparing metadata (setup.py): started
Preparing metadata (setup.py): finished with status 'done'
Requirement already satisfied: torch in c:\users\asus\anaconda3\lib\site-
packages (from efficientnet_pytorch) (2.1.0+cu118)
Requirement already satisfied: filelock in c:\users\asus\anaconda3\lib\site-
packages (from torch->efficientnet_pytorch) (3.6.0)
Requirement already satisfied: typing-extensions in
c:\users\asus\anaconda3\lib\site-packages (from torch->efficientnet_pytorch)
(4.7.1)
Requirement already satisfied: sympy in c:\users\asus\anaconda3\lib\site-
Requirement already satisfied: networkx in c:\users\asus\anaconda3\lib\site-
Requirement already satisfied: jinja2 in c:\users\asus\anaconda3\lib\site-
Requirement already satisfied: fsspec in c:\users\asus\anaconda3\lib\site-
22
Requirement already satisfied: MarkupSafe>=2.0 in
c:\users\asus\anaconda3\lib\site-packages (from
jinja2->torch->efficientnet_pytorch) (2.0.1)
Requirement already satisfied: mpmath>=0.19 in c:\users\asus\anaconda3\lib\site-
packages (from sympy->torch->efficientnet_pytorch) (1.2.1)
Building wheels for collected packages: efficientnet_pytorch
Building wheel for efficientnet_pytorch (setup.py): started
Building wheel for efficientnet_pytorch (setup.py): finished with status
'done'
Created wheel for efficientnet_pytorch:
filename=efficientnet_pytorch-0.7.1-py3-none-any.whl size=16428
sha256=7e66585381237a390daa1f439fd5ee3f06ee24e47402634d0b009bfb9c1d7e10
Stored in directory: c:\users\asus\appdata\local\pip\cache\wheels\29\16\24\752
e89d88d333af39a288421e64d613b5f652918e39ef1f8e3
Successfully built efficientnet_pytorch
Installing collected packages: efficientnet_pytorch
Successfully installed efficientnet_pytorch-0.7.1
0.3.1 1 and 2. Initialise the Training Model and Load Data

Here, we are using the state of the art EﬀicientNey Architecture
[68]: import torch

import torch.optim as optim
from torchvision import datasets, transforms
from efficientnet_pytorch import EfficientNet
import matplotlib.pyplot as plt
# Set device (GPU/CPU)

# Define transforms
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
# Load datasets
train_dataset = datasets.ImageFolder('train/', transform=transform)
test_dataset = datasets.ImageFolder('test/', transform=transform)

train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=32,␣
↪shuffle=True)
23
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=32)
# Load pre-trained EfficientNet with ImageNet weights

efficientnet = EfficientNet.from_pretrained('efficientnet-b0',␣
↪num_classes=len(train_dataset.classes))
# Modify the model to fit the desired number of parameters

# For example, reducing the number of parameters to 10M
# Modify model architecture accordingly to fit the constraint
# Define criterion and optimizer

optimizer = optim.Adam(efficientnet.parameters(), lr=0.001)
Loaded pretrained weights for efficientnet-b0
[69]: efficientnet
[69]: EfficientNet(
(_conv_stem): Conv2dStaticSamePadding(
3, 32, kernel_size=(3, 3), stride=(2, 2), bias=False
(static_padding): ZeroPad2d((0, 1, 0, 1))
)
(_bn0): BatchNorm2d(32, eps=0.001, momentum=0.010000000000000009, affine=True,
(_blocks): ModuleList(
(0): MBConvBlock(
(_depthwise_conv): Conv2dStaticSamePadding(
32, 32, kernel_size=(3, 3), stride=[1, 1], groups=32, bias=False
)
(_bn1): BatchNorm2d(32, eps=0.001, momentum=0.010000000000000009,
affine=True, track_running_stats=True)
(_se_reduce): Conv2dStaticSamePadding(
32, 8, kernel_size=(1, 1), stride=(1, 1)
(static_padding): Identity()
)
(_se_expand): Conv2dStaticSamePadding(
)
(_project_conv): Conv2dStaticSamePadding(
)
24
(_swish): MemoryEfficientSwish()
)
(1): MBConvBlock(
(_expand_conv): Conv2dStaticSamePadding(
)
)
)
)
)
)
(2): MBConvBlock(
)
144, 144, kernel_size=(3, 3), stride=(1, 1), groups=144, bias=False
)
25
)
)
)
)
(3): MBConvBlock(
)
)
)
)
)
)
(4): MBConvBlock(
)
26
)
)
)
)
)
(5): MBConvBlock(
)
)
)
)
27
)
)
(6-7): 2 x MBConvBlock(
)
)
)
)
)
)
(8): MBConvBlock(
)
)
28
)
)
)
)
)
)
)
)
)
)
(11): MBConvBlock(
29
)
)
)
)
)
)
)
)
)
30
)
)
)
(15): MBConvBlock(
)
)
)
)
)
)
)
(_conv_head): Conv2dStaticSamePadding(
)
(_avg_pooling): AdaptiveAvgPool2d(output_size=1)
31
(_dropout): Dropout(p=0.2, inplace=False)
(_fc): Linear(in_features=1280, out_features=200, bias=True)
)
[ ]:
[70]: # Training loop

def train_efficient(model, train_loader, criterion, optimizer, num_epochs=10):
model.to(device)
model.train()
train_losses = []
train_accuracy = []

running_loss = 0.0
correct = 0
total = 0

loss.backward()
optimizer.step()


Accuracy: {epoch_accuracy:.2f}%")
↪
32
# Train the model
train_losses_e, train_accuracy_e = train_efficient(efficientnet, train_loader,␣
↪criterion, optimizer)

[71]: # Plot training loss and accuracy curves

plt.plot(train_losses_e, label='Training Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title('Training Loss Curve')
plt.legend()
plt.plot(train_accuracy_e, label='Training Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy (%)')
plt.title('Training Accuracy Curve')
plt.legend()
plt.show()
33
[72]: # Evaluate the model on the test set
def evaluate_efficient(model, test_loader):
model.eval()
correct = 0
total = 0
test_accuracy = 100. * correct / total

print(f"Test Accuracy: {test_accuracy:.2f}%")
# Evaluate the model

evaluate_efficient(efficientnet, test_loader)
# Save the final model checkpoint

torch.save(efficientnet.state_dict(), 'efficientnet_checkpoint.pth')
Test Accuracy: 58.27%
34
[ ]:
[73]: model = efficientnet
0.4 Resnet
[ ]:
[82]: import torch

import torchvision.transforms as transforms
import torch.optim as optim
import torchvision.datasets as datasets
import matplotlib.pyplot as plt
0.4.1 1 and 2. Initialise the Training Model and Load Data

Here, we are using the age old good Resnet Model
[84]: # Define data transforms

transforms.Resize(256),
transforms.CenterCrop(224),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
# Load dataset
train_dataset = datasets.ImageFolder('train', transform=transform)
test_dataset = datasets.ImageFolder('test', transform=transform)

train_loader_resnet = DataLoader(train_dataset, batch_size=64, shuffle=True)
test_loader_resnet = DataLoader(test_dataset, batch_size=64, shuffle=False)
# Initialize the pretrained ResNet model

model_resnet = models.resnet18(pretrained=True)
# Modify the fully connected layer to match the number of classes in your␣
↪dataset
model_resnet.fc = nn.Linear(model_resnet.fc.in_features, num_classes)
# Define loss function and optimizer

optimizer = optim.Adam(model_resnet.parameters(), lr=0.001)
35
0.4.2 3. Define the Training Loop and also 5. Evaluating on the Go
[85]: # Training function
def train_resnet(model, train_loader, criterion, optimizer, num_epochs=10):
train_losses_resnet = []
train_accuracies_resnet = []
test_losses_resnet = []
test_accuracies_resnet = []

model.train()
running_loss = 0.0
correct = 0
total = 0
for images, labels in train_loader:
outputs = model(images)
loss.backward()
optimizer.step()
train_loss = running_loss / len(train_loader)

train_accuracy = correct / total
train_losses_resnet.append(train_loss)
train_accuracies_resnet.append(train_accuracy)
# Evaluation on test set

model.eval()
test_loss = 0.0
correct = 0
total = 0
for images, labels in test_loader:
outputs = model(images)
test_loss += loss.item()
test_loss /= len(test_loader)
test_accuracy = correct / total
36
test_losses_resnet.append(test_loss)
test_accuracies_resnet.append(test_accuracy)
print(f'Epoch {epoch+1}/{num_epochs}, '

f'Train Loss: {train_loss:.4f}, Train Acc: {train_accuracy:.4f}, '
f'Test Loss: {test_loss:.4f}, Test Acc: {test_accuracy:.4f}')
return train_losses_resnet, train_accuracies_resnet, test_losses_resnet,␣

↪test_accuracies_resnet
[86]: # Train the model

train_losses_resnet, train_accuracies_resnet, test_losses_resnet,␣
↪test_accuracies_resnet = train_resnet(model_resnet, train_loader_resnet,␣
↪criterion, optimizer)
# Save the trained model checkpoint

torch.save(model_resnet.state_dict(), 'pretrained_resnet_checkpoint.pth')
Epoch 1/10, Train Loss: 3.8045, Train Acc: 0.1775, Test Loss: 2.9566, Test Acc:
0.2639
0.3650
0.4693
0.5133
0.5773
0.6044
0.6624
0.6707
0.6907
0.6919

[87]: # Plotting training curves
plt.plot(train_losses_resnet, label='Train Loss')
plt.plot(test_losses_resnet, label='Test Loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
37
plt.title('Training and Test Loss')
plt.legend()
plt.show()
plt.plot(train_accuracies_resnet, label='Train Accuracy')
plt.plot(test_accuracies_resnet, label='Test Accuracy')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.title('Training and Test Accuracy')
plt.legend()
plt.show()
38
[121]: import torch
from torchsummary import summary
from graphviz import Digraph
# Print the summary of the model

summary(efficientnet, (3, 224, 224))
# Create a PyTorch summary

summary_str = str(summary(efficientnet, (3, 224, 224)))
# Function to convert the PyTorch summary to a graph

def torchsummary_to_graph(summary_str):
# Initialize graph
dot = Digraph()
# Add nodes and edges to the graph

layers = summary_str.split('\n')[2:]
for layer in layers:
layer_info = layer.split()
layer_name = layer_info[0]
input_shape = layer_info[1]
output_shape = layer_info[2]
params = layer_info[3]
39
dot.node(layer_name, label=f"{layer_name}\n{input_shape} ->␣
↪{output_shape}\nParams: {params}")
if 'Conv' in layer_name:
dot.attr('node', shape='rectangle')
for i in range(len(layers) - 1):

dot.edge(layers[i].split()[0], layers[i + 1].split()[0])
return dot
# Create the graph

graph = torchsummary_to_graph(summary_str)
# Save the graph as an image file

graph.render('model_resnet_architecture', format='png', cleanup=True)
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
ZeroPad2d-1 [-1, 3, 225, 225] 0
Conv2dStaticSamePadding-2 [-1, 32, 112, 112] 864
BatchNorm2d-3 [-1, 32, 112, 112] 64
MemoryEfficientSwish-4 [-1, 32, 112, 112] 0
ZeroPad2d-5 [-1, 32, 114, 114] 0
BatchNorm2d-7 [-1, 32, 112, 112] 64
Identity-9 [-1, 32, 1, 1] 0
Identity-12 [-1, 8, 1, 1] 0
Identity-14 [-1, 32, 112, 112] 0
BatchNorm2d-16 [-1, 16, 112, 112] 32
MBConvBlock-17 [-1, 16, 112, 112] 0
Identity-18 [-1, 16, 112, 112] 0
Conv2dStaticSamePadding-19 [-1, 96, 112, 112] 1,536
BatchNorm2d-20 [-1, 96, 112, 112] 192
ZeroPad2d-22 [-1, 96, 113, 113] 0
BatchNorm2d-24 [-1, 96, 56, 56] 192
Identity-26 [-1, 96, 1, 1] 0
40
Identity-29 [-1, 4, 1, 1] 0
Identity-31 [-1, 96, 56, 56] 0
BatchNorm2d-33 [-1, 24, 56, 56] 48
MBConvBlock-34 [-1, 24, 56, 56] 0
Identity-35 [-1, 24, 56, 56] 0
BatchNorm2d-37 [-1, 144, 56, 56] 288
ZeroPad2d-39 [-1, 144, 58, 58] 0
BatchNorm2d-41 [-1, 144, 56, 56] 288
Identity-43 [-1, 144, 1, 1] 0
Identity-46 [-1, 6, 1, 1] 0
Identity-48 [-1, 144, 56, 56] 0
BatchNorm2d-50 [-1, 24, 56, 56] 48
MBConvBlock-51 [-1, 24, 56, 56] 0
Identity-52 [-1, 24, 56, 56] 0
BatchNorm2d-54 [-1, 144, 56, 56] 288
ZeroPad2d-56 [-1, 144, 59, 59] 0
BatchNorm2d-58 [-1, 144, 28, 28] 288
Identity-60 [-1, 144, 1, 1] 0
Identity-63 [-1, 6, 1, 1] 0
Identity-65 [-1, 144, 28, 28] 0
BatchNorm2d-67 [-1, 40, 28, 28] 80
MBConvBlock-68 [-1, 40, 28, 28] 0
Identity-69 [-1, 40, 28, 28] 0
BatchNorm2d-71 [-1, 240, 28, 28] 480
ZeroPad2d-73 [-1, 240, 32, 32] 0
BatchNorm2d-75 [-1, 240, 28, 28] 480
41
Identity-77 [-1, 240, 1, 1] 0
Identity-80 [-1, 10, 1, 1] 0
Identity-82 [-1, 240, 28, 28] 0
BatchNorm2d-84 [-1, 40, 28, 28] 80
MBConvBlock-85 [-1, 40, 28, 28] 0
Identity-86 [-1, 40, 28, 28] 0
BatchNorm2d-88 [-1, 240, 28, 28] 480
ZeroPad2d-90 [-1, 240, 29, 29] 0
BatchNorm2d-92 [-1, 240, 14, 14] 480
Identity-94 [-1, 240, 1, 1] 0
Identity-97 [-1, 10, 1, 1] 0
Identity-99 [-1, 240, 14, 14] 0
BatchNorm2d-101 [-1, 80, 14, 14] 160
MBConvBlock-102 [-1, 80, 14, 14] 0
Identity-103 [-1, 80, 14, 14] 0
BatchNorm2d-105 [-1, 480, 14, 14] 960
ZeroPad2d-107 [-1, 480, 16, 16] 0
BatchNorm2d-109 [-1, 480, 14, 14] 960
Identity-111 [-1, 480, 1, 1] 0
Identity-114 [-1, 20, 1, 1] 0
Identity-116 [-1, 480, 14, 14] 0
BatchNorm2d-118 [-1, 80, 14, 14] 160
MBConvBlock-119 [-1, 80, 14, 14] 0
Identity-120 [-1, 80, 14, 14] 0
BatchNorm2d-122 [-1, 480, 14, 14] 960
ZeroPad2d-124 [-1, 480, 16, 16] 0
42
BatchNorm2d-126 [-1, 480, 14, 14] 960
Identity-128 [-1, 480, 1, 1] 0
Identity-131 [-1, 20, 1, 1] 0
Identity-133 [-1, 480, 14, 14] 0
BatchNorm2d-135 [-1, 80, 14, 14] 160
MBConvBlock-136 [-1, 80, 14, 14] 0
Identity-137 [-1, 80, 14, 14] 0
BatchNorm2d-139 [-1, 480, 14, 14] 960
ZeroPad2d-141 [-1, 480, 18, 18] 0
BatchNorm2d-143 [-1, 480, 14, 14] 960
Identity-145 [-1, 480, 1, 1] 0
Identity-148 [-1, 20, 1, 1] 0
Identity-150 [-1, 480, 14, 14] 0
BatchNorm2d-152 [-1, 112, 14, 14] 224
MBConvBlock-153 [-1, 112, 14, 14] 0
Identity-154 [-1, 112, 14, 14] 0
BatchNorm2d-156 [-1, 672, 14, 14] 1,344
ZeroPad2d-158 [-1, 672, 18, 18] 0
BatchNorm2d-160 [-1, 672, 14, 14] 1,344
Identity-162 [-1, 672, 1, 1] 0
Identity-165 [-1, 28, 1, 1] 0
Identity-167 [-1, 672, 14, 14] 0
BatchNorm2d-169 [-1, 112, 14, 14] 224
MBConvBlock-170 [-1, 112, 14, 14] 0
Identity-171 [-1, 112, 14, 14] 0
43
BatchNorm2d-173 [-1, 672, 14, 14] 1,344
ZeroPad2d-175 [-1, 672, 18, 18] 0
BatchNorm2d-177 [-1, 672, 14, 14] 1,344
Identity-179 [-1, 672, 1, 1] 0
Identity-182 [-1, 28, 1, 1] 0
Identity-184 [-1, 672, 14, 14] 0
BatchNorm2d-186 [-1, 112, 14, 14] 224
MBConvBlock-187 [-1, 112, 14, 14] 0
Identity-188 [-1, 112, 14, 14] 0
BatchNorm2d-190 [-1, 672, 14, 14] 1,344
ZeroPad2d-192 [-1, 672, 17, 17] 0
BatchNorm2d-194 [-1, 672, 7, 7] 1,344
Identity-196 [-1, 672, 1, 1] 0
Identity-199 [-1, 28, 1, 1] 0
Identity-201 [-1, 672, 7, 7] 0
BatchNorm2d-203 [-1, 192, 7, 7] 384
MBConvBlock-204 [-1, 192, 7, 7] 0
Identity-205 [-1, 192, 7, 7] 0
BatchNorm2d-207 [-1, 1152, 7, 7] 2,304
ZeroPad2d-209 [-1, 1152, 11, 11] 0
BatchNorm2d-211 [-1, 1152, 7, 7] 2,304
Identity-213 [-1, 1152, 1, 1] 0
Identity-216 [-1, 48, 1, 1] 0
Identity-218 [-1, 1152, 7, 7] 0
BatchNorm2d-220 [-1, 192, 7, 7] 384
44
MBConvBlock-221 [-1, 192, 7, 7] 0
Identity-222 [-1, 192, 7, 7] 0
BatchNorm2d-224 [-1, 1152, 7, 7] 2,304
ZeroPad2d-226 [-1, 1152, 11, 11] 0
BatchNorm2d-228 [-1, 1152, 7, 7] 2,304
Identity-230 [-1, 1152, 1, 1] 0
Identity-233 [-1, 48, 1, 1] 0
Identity-235 [-1, 1152, 7, 7] 0
BatchNorm2d-237 [-1, 192, 7, 7] 384
MBConvBlock-238 [-1, 192, 7, 7] 0
Identity-239 [-1, 192, 7, 7] 0
BatchNorm2d-241 [-1, 1152, 7, 7] 2,304
ZeroPad2d-243 [-1, 1152, 11, 11] 0
BatchNorm2d-245 [-1, 1152, 7, 7] 2,304
Identity-247 [-1, 1152, 1, 1] 0
Identity-250 [-1, 48, 1, 1] 0
Identity-252 [-1, 1152, 7, 7] 0
BatchNorm2d-254 [-1, 192, 7, 7] 384
MBConvBlock-255 [-1, 192, 7, 7] 0
Identity-256 [-1, 192, 7, 7] 0
BatchNorm2d-258 [-1, 1152, 7, 7] 2,304
ZeroPad2d-260 [-1, 1152, 9, 9] 0
BatchNorm2d-262 [-1, 1152, 7, 7] 2,304
Identity-264 [-1, 1152, 1, 1] 0
Identity-267 [-1, 48, 1, 1] 0
45
Identity-269 [-1, 1152, 7, 7] 0
BatchNorm2d-271 [-1, 320, 7, 7] 640
MBConvBlock-272 [-1, 320, 7, 7] 0
Identity-273 [-1, 320, 7, 7] 0
BatchNorm2d-275 [-1, 1280, 7, 7] 2,560
AdaptiveAvgPool2d-277 [-1, 1280, 1, 1] 0
Dropout-278 [-1, 1280] 0
Linear-279 [-1, 200] 256,200
================================================================
Total params: 4,263,748
Trainable params: 4,263,748
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 212.79
Params size (MB): 16.26
Estimated Total Size (MB): 229.63
----------------------------------------------------------------
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
ZeroPad2d-1 [-1, 3, 225, 225] 0
BatchNorm2d-3 [-1, 32, 112, 112] 64
ZeroPad2d-5 [-1, 32, 114, 114] 0
BatchNorm2d-7 [-1, 32, 112, 112] 64
Identity-9 [-1, 32, 1, 1] 0
Identity-12 [-1, 8, 1, 1] 0
Identity-14 [-1, 32, 112, 112] 0
BatchNorm2d-16 [-1, 16, 112, 112] 32
MBConvBlock-17 [-1, 16, 112, 112] 0
Identity-18 [-1, 16, 112, 112] 0
BatchNorm2d-20 [-1, 96, 112, 112] 192
ZeroPad2d-22 [-1, 96, 113, 113] 0
BatchNorm2d-24 [-1, 96, 56, 56] 192
46
Identity-26 [-1, 96, 1, 1] 0
Identity-29 [-1, 4, 1, 1] 0
Identity-31 [-1, 96, 56, 56] 0
BatchNorm2d-33 [-1, 24, 56, 56] 48
MBConvBlock-34 [-1, 24, 56, 56] 0
Identity-35 [-1, 24, 56, 56] 0
BatchNorm2d-37 [-1, 144, 56, 56] 288
ZeroPad2d-39 [-1, 144, 58, 58] 0
BatchNorm2d-41 [-1, 144, 56, 56] 288
Identity-43 [-1, 144, 1, 1] 0
Identity-46 [-1, 6, 1, 1] 0
Identity-48 [-1, 144, 56, 56] 0
BatchNorm2d-50 [-1, 24, 56, 56] 48
MBConvBlock-51 [-1, 24, 56, 56] 0
Identity-52 [-1, 24, 56, 56] 0
BatchNorm2d-54 [-1, 144, 56, 56] 288
ZeroPad2d-56 [-1, 144, 59, 59] 0
BatchNorm2d-58 [-1, 144, 28, 28] 288
Identity-60 [-1, 144, 1, 1] 0
Identity-63 [-1, 6, 1, 1] 0
Identity-65 [-1, 144, 28, 28] 0
BatchNorm2d-67 [-1, 40, 28, 28] 80
MBConvBlock-68 [-1, 40, 28, 28] 0
Identity-69 [-1, 40, 28, 28] 0
BatchNorm2d-71 [-1, 240, 28, 28] 480
47
ZeroPad2d-73 [-1, 240, 32, 32] 0
BatchNorm2d-75 [-1, 240, 28, 28] 480
Identity-77 [-1, 240, 1, 1] 0
Identity-80 [-1, 10, 1, 1] 0
Identity-82 [-1, 240, 28, 28] 0
BatchNorm2d-84 [-1, 40, 28, 28] 80
MBConvBlock-85 [-1, 40, 28, 28] 0
Identity-86 [-1, 40, 28, 28] 0
BatchNorm2d-88 [-1, 240, 28, 28] 480
ZeroPad2d-90 [-1, 240, 29, 29] 0
BatchNorm2d-92 [-1, 240, 14, 14] 480
Identity-94 [-1, 240, 1, 1] 0
Identity-97 [-1, 10, 1, 1] 0
Identity-99 [-1, 240, 14, 14] 0
BatchNorm2d-101 [-1, 80, 14, 14] 160
MBConvBlock-102 [-1, 80, 14, 14] 0
Identity-103 [-1, 80, 14, 14] 0
BatchNorm2d-105 [-1, 480, 14, 14] 960
ZeroPad2d-107 [-1, 480, 16, 16] 0
BatchNorm2d-109 [-1, 480, 14, 14] 960
Identity-111 [-1, 480, 1, 1] 0
Identity-114 [-1, 20, 1, 1] 0
Identity-116 [-1, 480, 14, 14] 0
BatchNorm2d-118 [-1, 80, 14, 14] 160
MBConvBlock-119 [-1, 80, 14, 14] 0
Identity-120 [-1, 80, 14, 14] 0
48
BatchNorm2d-122 [-1, 480, 14, 14] 960
ZeroPad2d-124 [-1, 480, 16, 16] 0
BatchNorm2d-126 [-1, 480, 14, 14] 960
Identity-128 [-1, 480, 1, 1] 0
Identity-131 [-1, 20, 1, 1] 0
Identity-133 [-1, 480, 14, 14] 0
BatchNorm2d-135 [-1, 80, 14, 14] 160
MBConvBlock-136 [-1, 80, 14, 14] 0
Identity-137 [-1, 80, 14, 14] 0
BatchNorm2d-139 [-1, 480, 14, 14] 960
ZeroPad2d-141 [-1, 480, 18, 18] 0
BatchNorm2d-143 [-1, 480, 14, 14] 960
Identity-145 [-1, 480, 1, 1] 0
Identity-148 [-1, 20, 1, 1] 0
Identity-150 [-1, 480, 14, 14] 0
BatchNorm2d-152 [-1, 112, 14, 14] 224
MBConvBlock-153 [-1, 112, 14, 14] 0
Identity-154 [-1, 112, 14, 14] 0
BatchNorm2d-156 [-1, 672, 14, 14] 1,344
ZeroPad2d-158 [-1, 672, 18, 18] 0
BatchNorm2d-160 [-1, 672, 14, 14] 1,344
Identity-162 [-1, 672, 1, 1] 0
Identity-165 [-1, 28, 1, 1] 0
Identity-167 [-1, 672, 14, 14] 0
49
BatchNorm2d-169 [-1, 112, 14, 14] 224
MBConvBlock-170 [-1, 112, 14, 14] 0
Identity-171 [-1, 112, 14, 14] 0
BatchNorm2d-173 [-1, 672, 14, 14] 1,344
ZeroPad2d-175 [-1, 672, 18, 18] 0
BatchNorm2d-177 [-1, 672, 14, 14] 1,344
Identity-179 [-1, 672, 1, 1] 0
Identity-182 [-1, 28, 1, 1] 0
Identity-184 [-1, 672, 14, 14] 0
BatchNorm2d-186 [-1, 112, 14, 14] 224
MBConvBlock-187 [-1, 112, 14, 14] 0
Identity-188 [-1, 112, 14, 14] 0
BatchNorm2d-190 [-1, 672, 14, 14] 1,344
ZeroPad2d-192 [-1, 672, 17, 17] 0
BatchNorm2d-194 [-1, 672, 7, 7] 1,344
Identity-196 [-1, 672, 1, 1] 0
Identity-199 [-1, 28, 1, 1] 0
Identity-201 [-1, 672, 7, 7] 0
BatchNorm2d-203 [-1, 192, 7, 7] 384
MBConvBlock-204 [-1, 192, 7, 7] 0
Identity-205 [-1, 192, 7, 7] 0
BatchNorm2d-207 [-1, 1152, 7, 7] 2,304
ZeroPad2d-209 [-1, 1152, 11, 11] 0
BatchNorm2d-211 [-1, 1152, 7, 7] 2,304
Identity-213 [-1, 1152, 1, 1] 0
Identity-216 [-1, 48, 1, 1] 0
50
Identity-218 [-1, 1152, 7, 7] 0
BatchNorm2d-220 [-1, 192, 7, 7] 384
MBConvBlock-221 [-1, 192, 7, 7] 0
Identity-222 [-1, 192, 7, 7] 0
BatchNorm2d-224 [-1, 1152, 7, 7] 2,304
ZeroPad2d-226 [-1, 1152, 11, 11] 0
BatchNorm2d-228 [-1, 1152, 7, 7] 2,304
Identity-230 [-1, 1152, 1, 1] 0
Identity-233 [-1, 48, 1, 1] 0
Identity-235 [-1, 1152, 7, 7] 0
BatchNorm2d-237 [-1, 192, 7, 7] 384
MBConvBlock-238 [-1, 192, 7, 7] 0
Identity-239 [-1, 192, 7, 7] 0
BatchNorm2d-241 [-1, 1152, 7, 7] 2,304
ZeroPad2d-243 [-1, 1152, 11, 11] 0
BatchNorm2d-245 [-1, 1152, 7, 7] 2,304
Identity-247 [-1, 1152, 1, 1] 0
Identity-250 [-1, 48, 1, 1] 0
Identity-252 [-1, 1152, 7, 7] 0
BatchNorm2d-254 [-1, 192, 7, 7] 384
MBConvBlock-255 [-1, 192, 7, 7] 0
Identity-256 [-1, 192, 7, 7] 0
BatchNorm2d-258 [-1, 1152, 7, 7] 2,304
ZeroPad2d-260 [-1, 1152, 9, 9] 0
BatchNorm2d-262 [-1, 1152, 7, 7] 2,304
Identity-264 [-1, 1152, 1, 1] 0
51
Identity-267 [-1, 48, 1, 1] 0
Identity-269 [-1, 1152, 7, 7] 0
BatchNorm2d-271 [-1, 320, 7, 7] 640
MBConvBlock-272 [-1, 320, 7, 7] 0
Identity-273 [-1, 320, 7, 7] 0
BatchNorm2d-275 [-1, 1280, 7, 7] 2,560
AdaptiveAvgPool2d-277 [-1, 1280, 1, 1] 0
Dropout-278 [-1, 1280] 0
Linear-279 [-1, 200] 256,200
================================================================
Total params: 4,263,748
Trainable params: 4,263,748
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.57
Forward/backward pass size (MB): 212.79
Params size (MB): 16.26
Estimated Total Size (MB): 229.63
----------------------------------------------------------------
---------------------------------------------------------------------------
FileNotFoundError Traceback (most recent call last)
File ~\anaconda3\lib\site-packages\graphviz\backend\execute.py:81, in␣
↪run_check(cmd, input_lines, encoding, quiet, **kwargs)
80 else:
---> 81 proc = subprocess.run(cmd, **kwargs)
82 except OSError as e:
File ~\anaconda3\lib\subprocess.py:505, in run(input, capture_output, timeout,␣

↪check, *popenargs, **kwargs)
503 kwargs['stderr'] = PIPE

--> 505 with Popen(*popenargs, **kwargs) as process:
506 try:
File ~\anaconda3\lib\subprocess.py:951, in Popen.__init__(self, args, bufsize,␣

↪executable, stdin, stdout, stderr, preexec_fn, close_fds, shell, cwd, env,␣
↪universal_newlines, startupinfo, creationflags, restore_signals,␣
↪start_new_session, pass_fds, user, group, extra_groups, encoding, errors,␣
↪text, umask)
948 self.stderr = io.TextIOWrapper(self.stderr,

949 encoding=encoding, errors=errors)
--> 951 self._execute_child(args, executable, preexec_fn, close_fds,
52
952 pass_fds, cwd, env,
953 startupinfo, creationflags, shell,
954 p2cread, p2cwrite,
955 c2pread, c2pwrite,
956 errread, errwrite,
957 restore_signals,
958 gid, gids, uid, umask,
959 start_new_session)
960 except:
961 # Cleanup if the child failed starting.
File ~\anaconda3\lib\subprocess.py:1420, in Popen._execute_child(self, args,␣

↪executable, preexec_fn, close_fds, pass_fds, cwd, env, startupinfo,␣
↪creationflags, shell, p2cread, p2cwrite, c2pread, c2pwrite, errread, errwrite,␣
↪unused_restore_signals, unused_gid, unused_gids, unused_uid, unused_umask,␣
↪unused_start_new_session)
1419 try:
-> 1420 hp, ht, pid, tid = _winapi.CreateProcess(executable, args,
1421 # no special security
1422 None, None,
1423 int(not close_fds),
1424 creationflags,
1425 env,
1426 cwd,
1427 startupinfo)
1428 finally:
1429 # Child is launched. Close the parent's copy of those pipe
1430 # handles that only the child should have open. You need
(…)
1433 # pipe will not close when the child process exits and the
1434 # ReadFile will hang.
FileNotFoundError: [WinError 2] The system cannot find the file specified
The above exception was the direct cause of the following exception:
ExecutableNotFound Traceback (most recent call last)

Input In [121], in <cell line: 42>()
39 graph = torchsummary_to_graph(summary_str)
41 # Save the graph as an image file
---> 42 graph.render('model_resnet_architecture', format='png', cleanup=True)
File ~\anaconda3\lib\site-packages\graphviz\_tools.py:171, in␣

↪deprecate_positional_args.<locals>.decorator.<locals>.wrapper(*args, **kwargs)
162 wanted = ', '.join(f'{name}={value!r}'

163 for name, value in deprecated.items())
164 warnings.warn(f'The signature of {func.__name__} will be reduced'
165 f' to {supported_number} positional args'
53
166 f' {list(supported)}: pass {wanted}'
167 ' as keyword arg(s)',
168 stacklevel=stacklevel,
169 category=category)
--> 171 return func(*args, **kwargs)
File ~\anaconda3\lib\site-packages\graphviz\rendering.py:122, in Render.

↪render(self, filename, directory, view, cleanup, format, renderer, formatter,␣
↪neato_no_op, quiet, quiet_view, outfile, engine, raise_if_result_exists,␣
↪overwrite_source)
118 filepath = self.save(filename, directory=directory, skip_existing=None)

120 args.append(filepath)
--> 122 rendered = self._render(*args, **kwargs)
124 if cleanup:
125 log.debug('delete %r', filepath)
File ~\anaconda3\lib\site-packages\graphviz\_tools.py:171, in␣

↪deprecate_positional_args.<locals>.decorator.<locals>.wrapper(*args, **kwargs)
162 wanted = ', '.join(f'{name}={value!r}'

163 for name, value in deprecated.items())
164 warnings.warn(f'The signature of {func.__name__} will be reduced'
165 f' to {supported_number} positional args'
166 f' {list(supported)}: pass {wanted}'
167 ' as keyword arg(s)',
168 stacklevel=stacklevel,
169 category=category)
--> 171 return func(*args, **kwargs)
File ~\anaconda3\lib\site-packages\graphviz\backend\rendering.py:324, in␣

↪render(engine, format, filepath, renderer, formatter, neato_no_op, quiet,␣
↪outfile, raise_if_result_exists, overwrite_filepath)
320 raise exceptions.FileExistsError(f'output file exists: {os.

↪fspath(outfile)!r}')
322 cmd += args

--> 324 execute.run_check(cmd,
325 cwd=filepath.parent if filepath.parent.parts else None,
326 quiet=quiet,
327 capture_output=True)
329 return os.fspath(outfile)
File ~\anaconda3\lib\site-packages\graphviz\backend\execute.py:84, in␣

↪run_check(cmd, input_lines, encoding, quiet, **kwargs)
82 except OSError as e:
83 if e.errno == errno.ENOENT:
---> 84 raise ExecutableNotFound(cmd) from e
85 raise
87 if not quiet and proc.stderr:
54
ExecutableNotFound: failed to execute WindowsPath('dot'), make sure the Graphviz␣
↪executables are on your systems' PATH
[118]: # Move the model to GPU if available

device = torch.device("cuda")
model_resnet.to(device)
[118]: ResNet(
(conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3),
bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True,
(relu): ReLU(inplace=True)
(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1,
ceil_mode=False)
(layer1): Sequential(
(0): BasicBlock(
bias=False)
bias=False)
)
(1): BasicBlock(
bias=False)
bias=False)
)
)
(0): BasicBlock(
(conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1,
1), bias=False)
55
1), bias=False)
(downsample): Sequential(
)
)
(1): BasicBlock(
1), bias=False)
1), bias=False)
)
)
(0): BasicBlock(
1), bias=False)
1), bias=False)
)
)
(1): BasicBlock(
1), bias=False)
1), bias=False)
56
)
)
(0): BasicBlock(
1), bias=False)
1), bias=False)
)
)
(1): BasicBlock(
1), bias=False)
1), bias=False)
)
)
(avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
(fc): Linear(in_features=512, out_features=200, bias=True)
)
[ ]:
[109]: # Print the summary of the model

summary(model_resnet, (3, 224, 224))
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
Input In [109], in <cell line: 2>()
1 # Print the summary of the model
----> 2 summary(model_resnet, (3, 224, 224))
57
File ~\anaconda3\lib\site-packages\torchsummary\torchsummary.py:72, in␣
↪summary(model, input_size, batch_size, device)
68 model.apply(register_hook)
70 # make a forward pass
71 # print(x.shape)
---> 72 model(*x)
74 # remove these hooks
75 for h in hooks:
File ~\anaconda3\lib\site-packages\torch\nn\modules\module.py:1518, in Module.

↪_wrapped_call_impl(self, *args, **kwargs)
1516 return self._compiled_call_impl(*args, **kwargs) # type:␣

↪ignore[misc]
1517 else:
-> 1518 return self._call_impl(*args, **kwargs)

↪_call_impl(self, *args, **kwargs)
1522 # If we don't have any hooks, we want to skip the rest of the logic in
1523 # this function, and just call forward.
1524 if not (self._backward_hooks or self._backward_pre_hooks or self.
↪_forward_hooks or self._forward_pre_hooks
1525 or _global_backward_pre_hooks or _global_backward_hooks

1526 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1527 return forward_call(*args, **kwargs)
1529 try:
1530 result = None
File ~\anaconda3\lib\site-packages\torchvision\models\resnet.py:285, in ResNet.

↪forward(self, x)
284 def forward(self, x: Tensor) -> Tensor:

--> 285 return self._forward_impl(x)
File ~\anaconda3\lib\site-packages\torchvision\models\resnet.py:268, in ResNet.

↪_forward_impl(self, x)
266 def _forward_impl(self, x: Tensor) -> Tensor:

267 # See note [TorchScript super()]
--> 268 x = self.conv1(x)
269 x = self.bn1(x)
270 x = self.relu(x)

↪_wrapped_call_impl(self, *args, **kwargs)
1516 return self._compiled_call_impl(*args, **kwargs) # type:␣

↪ignore[misc]
1517 else:
-> 1518 return self._call_impl(*args, **kwargs)
58
↪_call_impl(self, *args, **kwargs)
1565 bw_hook = hooks.BackwardHook(self, full_backward_hooks,␣

↪backward_pre_hooks)
1566 args = bw_hook.setup_input_hook(args)

-> 1568 result = forward_call(*args, **kwargs)
1569 if _global_forward_hooks or self._forward_hooks:
1570 for hook_id, hook in (
1571 *_global_forward_hooks.items(),
1572 *self._forward_hooks.items(),
1573 ):
1574 # mark that always called hook is run
File ~\anaconda3\lib\site-packages\torch\nn\modules\conv.py:460, in Conv2d.

↪forward(self, input)
459 def forward(self, input: Tensor) -> Tensor:

--> 460 return self._conv_forward(input, self.weight, self.bias)
File ~\anaconda3\lib\site-packages\torch\nn\modules\conv.py:456, in Conv2d.

↪_conv_forward(self, input, weight, bias)
452 if self.padding_mode != 'zeros':

453 return F.conv2d(F.pad(input, self._reversed_padding_repeated_twice,␣
↪mode=self.padding_mode),
454 weight, bias, self.stride,

455 _pair(0), self.dilation, self.groups)
--> 456 return F.conv2d(input, weight, bias, self.stride,
457 self.padding, self.dilation, self.groups)
RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.

↪FloatTensor) should be the same
Total number of parameters

[98]: import torch
# Count the total number of parameters

total_params = sum(p.numel() for p in mobilenet.parameters())
print("Total number of parameters:", total_params)
Total number of parameters: 2480072
[99]: import torch

59
total_params = sum(p.numel() for p in efficientnet.parameters())
[100]: import torch


total_params = sum(p.numel() for p in inception_model.parameters())
[91]: !pip install graphviz
Collecting graphviz
Downloading graphviz-0.20.1-py3-none-any.whl.metadata (12 kB)
Downloading graphviz-0.20.1-py3-none-any.whl (47 kB)
---------------------------------------- 47.0/47.0 kB 2.3 MB/s eta 0:00:00
Installing collected packages: graphviz
Successfully installed graphviz-0.20.1
[101]: import torch


total_params = sum(p.numel() for p in model_resnet.parameters())

The weights of all are included in this drive link :
[ ]:
60

Project Mini

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Project Mini

Uploaded by

Copyright:

Available Formats

Project_Mini

# Define data transformations

# Create data loaders

[5]: import torch.nn as nn

# Initialize MobileNet model

# Freeze convolutional layers

# Unfreeze the last few layers for fine-tuning

0.1.3 3. Define the Training Loop

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

for epoch in range(num_epochs):

for inputs, labels in train_loader:

epoch_loss = running_loss / len(train_loader)

print(f"Epoch [{epoch+1}/{num_epochs}], Loss: {epoch_loss:.4f},␣

return train_losses, train_accuracy

train_losses_m, train_accuracy_m = train_m(mobilenet, train_loader, criterion,␣

Epoch [1/30], Loss: 0.5475, Accuracy: 88.81%

0.1.4 4. Plot the training results

[11]: import matplotlib.pyplot as plt

# Plot training details

[12]: torch.save(mobilenet.state_dict(), 'mobilenet_checkpoint.pth')

0.1.5 5. Evaluate the Model

accuracy = 100. * correct / total

Test Accuracy: 55.25%

0.2.1 1. Initialise the Training Model

[49]: import torch

# Load pre-trained Inception model

# Freeze convolutional layers

# Unfreeze the last few layers for fine-tuning

# Define transformations for data preprocessing

transforms.ToTensor(), # Convert images to tensors

# Load training and testing datasets

# Create data loaders

0.2.3 3. Define the Training Loop

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

def train(model, train_loader, criterion, optimizer, num_epochs=25):

for epoch in range(num_epochs):

for inputs, labels in train_loader:

# Resize input images to match Inception model's input size

# Get main output (logits) from the 'output' tensor

# Backward pass and optimization

epoch_loss = running_loss / len(train_loader)

print(f"Epoch [{epoch+1}/{num_epochs}], Loss: {epoch_loss:.4f},␣

return train_losses, train_accuracy

[59]: train(inception_model, train_loader, criterion, optimizer, num_epochs=20)

Epoch [1/20], Loss: 4.5101, Accuracy: 12.18%

[61]: torch.save(inception_model.state_dict(), 'inception_model.pth')

0.2.4 4. Evaluate the Model

accuracy = 100.*correct / total

Test Accuracy: 51.8813

0.2.5 5. Plot the training results

# Plot training loss and accuracy curves

[66]: !pip install efficientnet_pytorch

0.3.1 1 and 2. Initialise the Training Model and Load Data

[68]: import torch

# Set device (GPU/CPU)

# Create data loaders

# Load pre-trained EfficientNet with ImageNet weights

# Modify the model to fit the desired number of parameters

# Define criterion and optimizer

Loaded pretrained weights for efficientnet-b0

0.3.2 3. Define the Training Loop

[70]: # Training loop

File ~\anaconda3\lib\subprocess.py:951, in Popen.init(self, args, bufsize,␣