Built a system that harnesses the capabilities of transfer learning with the ResNet-50 architecture and the Annoy library to optimize the K-Nearest Neighbors (KNN) algorithm. By extracting features from over 30,000+ images using ResNet-50, the system can analyze and understand the visual data effectively. The recommendation process uses KNN to perform a similarity search, identifying the top 5 closest matches to a user's input and delivering personalized fashion suggestions.
This system is designed to be intuitive and efficient, showcasing the versatility of transfer learning, similarity search, and convolutional neural networks (CNNs). It serves as a solid platform for creating more advanced and extensive recommendation systems in the future.