Image datasets (CIFAR-10, ImageNet) in Computer Vision

Introduction

Image datasets help computers learn to recognize pictures by showing many examples. CIFAR-10 and ImageNet are popular collections of images used to teach machines about different objects.

When you want to train a computer to recognize everyday objects like cars, animals, or planes.

When testing how well a new image recognition method works on standard pictures.

When learning about machine learning with images and need a ready set of labeled photos.

When comparing different models to see which one identifies images more accurately.

When building apps that need to understand pictures, like photo search or self-driving cars.

Syntax

Computer Vision

from torchvision.datasets import CIFAR10

# Load CIFAR-10 dataset
cifar10_train = CIFAR10(root='./data', train=True, download=True)

# Load ImageNet dataset (requires manual download and setup)
# from torchvision.datasets import ImageNet
# imagenet_train = ImageNet(root='./data/imagenet', split='train')

CIFAR-10 is small and easy to download, with 60,000 images in 10 classes.

ImageNet is very large with over 1 million images in 1000 classes and needs manual setup.

Examples

This loads CIFAR-10 training images and prints how many images are in the set.

Computer Vision

from torchvision.datasets import CIFAR10

# Download and load training data
train_data = CIFAR10(root='./data', train=True, download=True)

# Check number of images
print(len(train_data))

This loads CIFAR-10 test images and shows the label of the first image.

Computer Vision

from torchvision.datasets import CIFAR10

# Load test data
test_data = CIFAR10(root='./data', train=False, download=True)

# Get first image and label
image, label = test_data[0]
print(f'Label: {label}')

Sample Model

This program loads CIFAR-10 images, trains a simple model for one batch, and checks accuracy on one batch of test images.

Computer Vision

import torch
from torchvision.datasets import CIFAR10
from torchvision.transforms import ToTensor
from torch.utils.data import DataLoader
import torchvision.models as models
import torch.nn as nn
import torch.optim as optim

# Load CIFAR-10 training and test data with transform to tensor
train_data = CIFAR10(root='./data', train=True, download=True, transform=ToTensor())
test_data = CIFAR10(root='./data', train=False, download=True, transform=ToTensor())

# Data loaders
train_loader = DataLoader(train_data, batch_size=64, shuffle=True)
test_loader = DataLoader(test_data, batch_size=64, shuffle=False)

# Use a simple pretrained model (ResNet18) and adjust for 10 classes
model = models.resnet18(pretrained=False)
model.fc = nn.Linear(model.fc.in_features, 10)

# Loss and optimizer
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.01)

# Train for 1 epoch (for demo)
model.train()
for images, labels in train_loader:
    optimizer.zero_grad()
    outputs = model(images)
    loss = criterion(outputs, labels)
    loss.backward()
    optimizer.step()
    break  # only 1 batch for quick demo

# Evaluate on test data
model.eval()
correct = 0
total = 0
with torch.no_grad():
    for images, labels in test_loader:
        outputs = model(images)
        _, predicted = torch.max(outputs, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()
        break  # only 1 batch for quick demo

accuracy = 100 * correct / total
print(f'Accuracy on 1 batch of test images: {accuracy:.2f}%')

Important Notes

CIFAR-10 images are small (32x32 pixels), so models train quickly but with limited detail.

ImageNet is much bigger and better for advanced models but needs more computing power.

Always check if datasets need manual download or special setup before use.

Summary

CIFAR-10 and ImageNet are popular image datasets for teaching computers to recognize pictures.

CIFAR-10 is small and easy to use; ImageNet is large and detailed.

These datasets help test and compare image recognition models in machine learning.