PFRS: Personalized Fashion Recommendation System Using EfficientNet

The fashion industry has undergone significant transformation owing to technological advancements, particularly in the realm of fashion recommendation systems. This paper introduces a novel approach, the Personalized Fashion Recommendation System (PFRS), which leverages reversed image processing, an EfficientNet pre-trained model, and the K-Nearest Neighbors (KNN) algorithm. Convolutional Neural Networks (ConvNets) are tailored within resource constraints and subsequently expanded to boost accuracy for PFRS. Notably, E-commerce platforms like Stitch Fix and ASOS employ Collaborative Filtering and Feedback Loop techniques, which rely on user-item interactions. Our research demonstrates that EfficientNet outperforms other transfer learning networks for object detection tasks, achieving 84.3% accuracy on the ImageNet dataset. Moreover, our EfficientNet exhibits an 8.4x reduction in size and a 6.1x increase in inference speed compared to leading ConvNet models like ResNet50 and Xception. This work lays a foundation for advanced fashion recommendation systems, enhancing user experience and efficiency in online fashion retail.