Personalized Interior Design Recommendation and Generation Using Transfer Learning

Abstract

Interior designing has become an integral part of designing a cozy,  practical,  and has an appealing space to live with.  Conventional interior designing requires the use of experienced designing experts which makes the whole process expensive and lengthy making it inaccessible to many users.  Therefore, in this paper, we present a Personalized Interior Design Recommendation and Generation using Transfer Learning framework,  that makes use of Artificial Intelligence and Computer vision to assist in generating personalized interior designing suggestions to the users. We employ a pre-trained Vision Transformer - Base/16( ViT-B/16) to categorize the uploaded image of room into different room categories like study room, living room, etc., and different design styles like modern, rustic, traditional, contemporary etc.  Post identification of the interior style, preferences like room size, budget range, furniture style, wall colour, flooring material, lighting, storage space needs, intelligent automation choices, eco-friendly appliances, luxury level, etc., are entered by the user to conjure the visual layout of an user-specified space with AI prompts.  Using transfer learning allows us to reach the desired accuracy levels much faster and with less computational complexity.  In conclusion, our framework provides an easy, inexpensive, and hassle-free personalized interior designing solution from the comfort of your house,  suiting the needs of homeowners, architects, interior designers and property agents.