I built a CIFAR-100 image classifier using a ResNet-50 architecture in PyTorch. The repository includes clean, reproducible training code, professional logging, and a Streamlit demo for interactive testing.

Key features include:

- Advanced data augmentation (Mixup, CutMix, AutoAugment)

- Optimized schedulers (CosineAnnealing, OneCycleLR)

- Regularization techniques (Label Smoothing, Gradient Clipping)

- Full training logs with accuracy/loss charts

The model achieves 84% accuracy, surpassing standard baselines. Feedback and technical suggestions are highly appreciated.