A Deep Learning Approach for Interior Design Style Classification and Idea Generation

This paper introduces an AI-based system for classifying interior design styles and providing personalized design suggestions using deep learning and web technologies. The project solves the dilemma of most users in defining and communicating their design taste due to limited knowledge and paralyzing options. Utilizing Convolutional Neural Networks (CNNs), more precisely the InceptionV3 model with transfer learning, the system categorizes pictures of a room into one of twelve pre-defined styles, i.e., Modern, Bohemian, Minimalist, and Traditional. The process consists of automated image scraping, removal of duplicates via perceptual hashing, and feature extraction via InceptionV3. A fully connected neural network classifier is trained on extracted features to make design style predictions. The model is incorporated in a Flask-based web application with the ability for users to upload room images, get style predictions, and see curated recommendations for furniture, color schemes, and layouts. The site also includes user registration, authentication, and a feedback loop for enhancing the quality of recommendations over time. This work illustrates how artificial intelligence can make the design ideation process easy and facilitate interior design for non-experts. Through the integration of CNN-based image classification with an interactive web interface, the system fosters creativity, personalization, and productivity, providing a scalable platform for intelligent design assistance in real-world applications.