With the exponential growth of e-commerce in the fashion industry, the need for efficient and personalized recommendation systems has become paramount. Existing systems often struggle to provide accurate suggestions tailored to individual preferences due to limitations in capturing fashion features and user preferences. In response to this challenge, we suggest a transfer learning model i.e. ResNet-50 and K-Nearest Neighbors (KNN) for application in a fashion recommendation system. ResNet-50 can encode a rich representation of clothes and accessories because it has been transferring learning optimized on a range of fashion datasets. Furthermore, based on user preferences, we rapidly obtain the fashion products that are most relevant by incorporating the KNN. Through thorough validation of real-world datasets, we demonstrate that our proposed method achieved 98% accuracy with 0.04 loss. All things considered, our invention represents a significant development in fashion recommendation systems and aims to revolutionize the online fashion buying experience.