CIFAR-10 Image Classification (PyTorch)

This repository implements image classification on the CIFAR-10 dataset using PyTorch. It includes a compact CNN baseline and a modular training/evaluation pipeline with augmentation, learning curves, and a confusion matrix.

---

✨ Features

• Dataset & Augmentation
  • torchvision.datasets.CIFAR10
  • Normalization (0.4914, 0.4822, 0.4465) / (0.2023, 0.1994, 0.2010)
  • RandomCrop(32, padding=4) and RandomHorizontalFlip() for training
• Model
  • 3-block CNN (Conv → BatchNorm → ReLU → MaxPool) with channels 32 → 64 → 128
  • Linear head: 128 × 4 × 4 → 10
• Training
  • Adam optimizer, CrossEntropyLoss
  • Configurable --epochs, --batch-size, --lr
• Evaluation & Visualization
  • Accuracy per epoch, final test accuracy
  • Classification report & confusion matrix
  • Loss/accuracy curves saved to results/