v1.0.0 — Dog Breed Classification

Ensemble transfer learning system for classifying dog breeds with 95% accuracy.

• ResNet50, EfficientNet B0, DenseNet121 transfer learning across 70 dog breeds
• Ensemble + test-time augmentation (TTA) achieves 95.29% accuracy (93.14% baseline)
• 7,946 training images; handles class imbalance (3.05 ratio) via weighted loss
• Inference: 217.53 FPS single model, 9.05 FPS with TTA