Project: Train a ResNet Vision Classifier with PyTorch

Capstone Objective

In this project, you will develop a complete end-to-end computer vision pipeline using PyTorch.

Requirements

1. Data Pipeline:
   • Implement random cropping, horizontal flips, and color jitter augmentations.
   • Use DataLoader with pinned memory and multi-worker prefetching.
2. Model Architecture:
   • Construct a modular ResidualBlock class.
   • Chain 4 residual stages with projection shortcuts when spatial dimensions downsample.
3. Training & Optimization:
   • Optimize using AdamW with cosine annealing learning rate scheduling.
   • Use torch.cuda.amp.autocast for FP16 mixed precision acceleration.
4. Deliverables:
   • Submit your GitHub repository URL or Colab link containing the complete code and evaluation plots.