Why custom data pipelines handle real data in PyTorch - Experiment to Prove It

Experiment - Why custom data pipelines handle real data

Problem:You want to train a model on real-world images, but the built-in data loaders do not fit your data format or preprocessing needs.

Current Metrics:Training accuracy: 90%, Validation accuracy: 60%

Issue:The model overfits because the data pipeline does not properly preprocess or augment the real data, causing poor validation performance.

Your Task

Create a custom PyTorch data pipeline that correctly loads, preprocesses, and augments real images to reduce overfitting and improve validation accuracy to above 75%.

Use PyTorch Dataset and DataLoader classes.

Do not change the model architecture.

Keep batch size and optimizer settings the same.

Hint 1

Hint 2

Hint 3

Hint 4

Solution

PyTorch

import torch
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms
from PIL import Image
import os

class CustomImageDataset(Dataset):
    def __init__(self, img_dir, labels_file, transform=None):
        self.img_dir = img_dir
        self.transform = transform
        with open(labels_file, 'r') as f:
            lines = f.readlines()
        self.img_labels = [line.strip().split() for line in lines]

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

    def __getitem__(self, idx):
        img_path = os.path.join(self.img_dir, self.img_labels[idx][0])
        image = Image.open(img_path).convert('RGB')
        label = int(self.img_labels[idx][1])
        if self.transform:
            image = self.transform(image)
        return image, label

# Define transforms including augmentation
transform = transforms.Compose([
    transforms.Resize((64, 64)),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406],
                         std=[0.229, 0.224, 0.225])
])

# Create dataset and dataloader
train_dataset = CustomImageDataset(img_dir='train_images', labels_file='train_labels.txt', transform=transform)
train_loader = DataLoader(train_dataset, batch_size=32, shuffle=True)

# Example training loop snippet
# for images, labels in train_loader:
#     outputs = model(images)
#     loss = criterion(outputs, labels)
#     optimizer.zero_grad()
#     loss.backward()
#     optimizer.step()

Created a custom Dataset class to load images and labels from files.

Added image resizing, normalization, and random horizontal flip for augmentation.

Used DataLoader with shuffling to improve training data randomness.

Kept model and training parameters unchanged.

Results Interpretation

Before: Training accuracy 90%, Validation accuracy 60% (overfitting)

After: Training accuracy 88%, Validation accuracy 78% (better generalization)

Custom data pipelines let you properly prepare and augment real data, which helps the model learn better and reduces overfitting.

Bonus Experiment

Try adding more data augmentations like random rotations and color jitter to see if validation accuracy improves further.

💡 Hint

Use torchvision.transforms.RandomRotation and transforms.ColorJitter in the transform pipeline.

Practice

1. Why do we use custom data pipelines in PyTorch when working with real data?

easy

A. To automatically create the model architecture

B. To make the model run faster without changing data

C. To avoid writing any code for data handling

D. To organize, transform, and prepare data for model training

Solution

1. Step 1: Understand the role of data pipelines

   Data pipelines help load, transform, and batch data so models can learn efficiently.

2. Step 2: Identify the main purpose

   Custom pipelines organize and prepare data, which is essential for training models well.

3. Final Answer:

   To organize, transform, and prepare data for model training -> Option D

4. Quick Check:

   Data pipelines = prepare data [OK]

Hint: Remember: pipelines prepare data, not build models [OK]

Common Mistakes:

• Thinking pipelines speed up model without data changes
• Believing pipelines create model architecture
• Assuming pipelines remove need for coding

2. Which PyTorch class is commonly extended to create a custom data pipeline?

easy

A. torch.utils.data.Dataset

B. torch.nn.Module

C. torch.optim.Optimizer

D. torch.Tensor

Solution

1. Step 1: Recall PyTorch data handling classes

   Custom data pipelines usually extend Dataset to define how data is loaded and accessed.

2. Step 2: Match class to purpose

   Dataset is designed for data loading; Module is for models, Optimizer for training steps, Tensor for data storage.

3. Final Answer:

   torch.utils.data.Dataset -> Option A

4. Quick Check:

   Custom pipeline = Dataset subclass [OK]

Hint: Dataset class is for data, Module is for models [OK]

Common Mistakes:

• Confusing Dataset with Module
• Choosing Optimizer instead of Dataset
• Picking Tensor as pipeline class

3. Given this custom Dataset code snippet, what will print(len(dataset)) output?

class MyDataset(torch.utils.data.Dataset):
    def __init__(self, data):
        self.data = data
    def __len__(self):
        return len(self.data)
    def __getitem__(self, idx):
        return self.data[idx]

data = [10, 20, 30, 40]
dataset = MyDataset(data)
print(len(dataset))

medium

A. 4

B. 3

C. Error: __len__ not defined

D. 0

Solution

1. Step 1: Check __len__ method

   The __len__ method returns the length of self.data, which is the list [10, 20, 30, 40].

2. Step 2: Calculate length of data list

   The list has 4 elements, so len(self.data) is 4.

3. Final Answer:

   4 -> Option A

4. Quick Check:

   Length of data list = 4 [OK]

Hint: len(dataset) calls __len__ which returns data length [OK]

Common Mistakes:

• Assuming __len__ is missing causing error
• Counting elements incorrectly
• Confusing __getitem__ with __len__

4. Identify the error in this custom Dataset code:

class MyDataset(torch.utils.data.Dataset):
    def __init__(self, data):
        self.data = data
    def __getitem__(self, idx):
        return self.data[idx]

# Missing __len__ method

medium

A. Missing __getitem__ method causes DataLoader to fail

B. DataLoader works fine without __len__

C. Missing __len__ method causes DataLoader to fail

D. No error, code is correct

Solution

1. Step 1: Check Dataset requirements

   PyTorch Dataset requires both __getitem__ and __len__ methods to work properly.

2. Step 2: Identify missing method

   The code lacks __len__, so DataLoader cannot know dataset size, causing failure.

3. Final Answer:

   Missing __len__ method causes DataLoader to fail -> Option C

4. Quick Check:

   Dataset needs __len__ and __getitem__ [OK]

Hint: Always define __len__ in custom Dataset [OK]

Common Mistakes:

• Thinking __getitem__ is missing
• Believing DataLoader works without __len__
• Ignoring Dataset method requirements

5. You have a dataset with images and labels stored in separate folders. Why is a custom data pipeline important here?

hard

A. To convert images into text files for easier reading

B. To load images and labels together, apply transforms, and batch them correctly

C. To automatically train the model without coding

D. To skip data preprocessing and feed raw files directly

Solution

1. Step 1: Understand data organization

   Images and labels in separate folders need to be paired correctly for training.

2. Step 2: Role of custom pipeline

   Custom pipeline loads both, applies necessary transforms (like resizing), and batches data for model input.

3. Final Answer:

   To load images and labels together, apply transforms, and batch them correctly -> Option B

4. Quick Check:

   Custom pipeline = organize and prepare complex data [OK]

Hint: Custom pipelines handle complex data loading and transforms [OK]

Common Mistakes:

• Assuming model trains without data preparation
• Thinking data converts to text automatically
• Skipping preprocessing steps