Deep Learning with Keras

By LiterallyTheOne

1: Load an Image Classification Dataset

Introduction

• Previous Tutorial: Select a Kaggle dataset
• This Tutorial: Load it correctly

Get a Dataset from Kaggle

• Kagglehub
• Download: top right of the page of the dataset

Kagglehub example

import kagglehub

# Download latest version
path = kagglehub.dataset_download("balabaskar/tom-and-jerry-image-classification")

print("Path to dataset files:", path)

Directory structure

class_a/
...a_image_1.jpg
...a_image_2.jpg
class_b/
...b_image_1.jpg
...b_image_2.jpg

Find the Image structure using ls

!ls {path}


"""
--------
output: 

challenges.csv   ground_truth.csv tom_and_jerry
"""

Moving forward

!ls {path}/tom_and_jerry


"""
--------
output: 

tom_and_jerry
"""

Find the directory that we want

!ls {path}/tom_and_jerry/tom_and_jerry


"""
--------
output: 

jerry       tom         tom_jerry_0 tom_jerry_1
"""

Store it in a variable

from pathlib import Path

data_path = Path(path) / "tom_and_jerry/tom_and_jerry"

ImageFolder

• Load an Image Classification Dataset
• Needs the structure below

main_directory/
...class_a/
......a_image_1.jpg
......a_image_2.jpg
...class_b/
......b_image_1.jpg
......b_image_2.jpg

ImageFolder example

from torchvision.datasets import ImageFolder

all_data = ImageFolder(data_path)

Show one sample

from matplotlib import pyplot as plt

for image, label in all_data:
    plt.figure()
    plt.imshow(image)
    print(label)
    break

"""
--------
output: 

0
"""

Transforms

• Transform images to our standard
• Torchvision

Resize Transform

from torchvision import transforms

# Without resize transform

all_data = ImageFolder(data_path)

for image, label in all_data:
    print(f"image size without resize transform: {image.size}")
    break

# With resize transform

transform = transforms.Resize((90, 160))

all_data = ImageFolder(data_path, transform=transform)

for image, label in all_data:
    print(f"image size with resize transform: {image.size}")
    break

"""
--------
output: 

image size without resize transform: (1280, 720)
image size with resize transform: (160, 90)
"""

ToTensor

• Changes PIL images to Tensor
• The way that model accepts our data

Combine two Transforms

trs = transforms.Compose(
    [
        transforms.Resize((90, 160)),
        transforms.ToTensor(),
    ]
)

all_data = ImageFolder(data_path, transform=trs)

for image, label in all_data:
    print(type(image))
    print(image.shape)
    break

"""
--------
output: 

<class 'torch.Tensor'>
torch.Size([3, 90, 160])
"""

Split into Train, Validation, and Test

• Some Datasets don't have these subsets
• We should make them manually
• random_split

Example of splitting

import torch
from torch.utils.data import random_split

g1 = torch.Generator().manual_seed(20)
train_data, val_data, test_data = random_split(all_data, [0.7, 0.2, 0.1], g1)

print(f"all_data's size: {len(all_data)}")
print(f"train_data's size: {len(train_data)}")
print(f"val_data's size: {len(val_data)}")
print(f"test_data's size: {len(test_data)}")

"""
--------
output: 

all_data's size: 5478
train_data's size: 3835
val_data's size: 1096
test_data's size: 547
"""

DataLoader

• Takes the loaded dataset
• Helps us to apply Deep Learning techniques
  • batch_size
  • shuffle

DataLoader for each subset

train_loader = DataLoader(train_data, batch_size=12, shuffle=True)
val_loader = DataLoader(val_data, batch_size=12, shuffle=False)
test_loader = DataLoader(test_data, batch_size=12, shuffle=False)

Show a batch of Data

fig, axes = plt.subplots(3, 4)

axes_ravel = axes.ravel()

for images, labels in train_loader:
    for i, (image, label) in enumerate(zip(images, labels)):
        axes_ravel[i].imshow(transforms.ToPILImage()(image))
        axes_ravel[i].set_axis_off()
        axes_ravel[i].set_title(f"{label}")
    break

Your Turn

• get your Kaggle dataset.
• use the ImageFolder to load that dataset and show one of its images.
• if you don’t have any of the train, validation, and test subsets, make them using random_split.
• load those three subsets using DataLoader and set a batch_size for them.
• show a batch of your data.

Link to the tutorial and materials