Knowledge Distillation for a Domain-Adaptive Visual Recommender System

In the last few years large-scale foundational models have shown remarkable performance in computer vision tasks. However, deploying such models in a production environment poses a significant challenge, because of their computational requirements. Furthermore, these models typically produce generic results and they often need some sort of external input. The concept of knowledge distillation provides a promising solution to this problem. By leveraging the teacher-student framework, the smaller”student” model learns to mimic the larger”teacher” model. In this paper, we focus on the challenges faced in the application of such techniques in the task of augmenting an object detection dataset used in a commercial Visual Recommender System that needs to detect items in various e-commerce websites, encompassing a wide range of product categories. We also present a simple solution to the problems we identified and propose a possible direction of future works.