Pokédex is an intelligent iOS application designed to identify physical Pokémon figures through Sim-to-Real Transfer Learning. The core challenge involved training a robust model solely on synthetic pixel art to recognize real-world 3D objects. I engineered the machine learning pipeline using TensorFlow and Keras, optimizing a MobileNetV2 architecture with advanced augmentation strategies—specifically random background compositing—to bridge this domain gap effectively.

The iOS client, built with SwiftUI and CoreML, executes real-time on-device inference without network dependencies. This project demonstrates the practical application of Computer Vision in mobile environments, addressing data scarcity through algorithmic efficiency rather than large datasets."

Disclaimer: Pokémon and Pokémon character names are trademarks of Nintendo. This is a non-commercial educational project.