This project applies transfer learning to classify images from the CIFAR-10 dataset, which contains 60,000 32×32 color images across 10 object categories (airplane, automobile, bird, cat, deer, dog, frog, horse, ship, truck).
Instead of training a deep model from scratch, I fine-tuned a ResNet18 pretrained on ImageNet. Transfer learning allows the model to leverage powerful feature extraction from large-scale datasets, making training faster and improving accuracy.
The project demonstrates how to:
- Load and preprocess CIFAR-10 images
- Apply transfer learning using PyTorch
- Fine-tune a pretrained model (ResNet18)
- Evaluate performance with accuracy and loss curves
- Python 3.10+
- PyTorch
- Torchvision
- ✅ Transfer learning with ResNet18
- ✅ CIFAR-10 dataset support
- ✅ GPU acceleration (CUDA)
- ✅ Training/validation loss & accuracy tracking
- Achieved high accuracy on CIFAR-10 using fine-tuned ResNet18
- Reduced training time compared to training from scratch
- Robust generalization on unseen test data