AI Knows Aesthetics: AI-Generated Interior Design Identification Using Deep Learning Algorithms

The integration of Artificial Intelligence (AI) in interior design has revolutionized how spaces are conceptualized, visualized, and classified. As digital tools and generative design models become increasingly sophisticated, distinguishing between AI-generated and real-world interior design images has emerged as a critical challenge. This classification is essential for ensuring authenticity in design visualization, improving recommendation systems, and enhancing virtual and augmented reality applications in architectural planning. Convolutional Neural Networks (CNNs) have proven to be highly effective in image classification tasks due to their ability to learn complex spatial hierarchies, making them particularly well-suited for differentiating between AI-rendered and real interior spaces. However, recent advancements in deep learning, including hybrid AI models and transformer-based vision networks, have further enhanced classification accuracy by integrating multimodal learning techniques. This study explores the use of advanced pre-trained CNN models, including DenseNet, RegNet, and SqueezeNet architecture, to classify interior design images. By leveraging a proprietary dataset composed of diverse AI-generated and real-world interior designs, our model achieved a peak classification accuracy of 97%, demonstrating superior performance compared to traditional deep learning techniques. Additionally, we introduce a novel image preprocessing technique that enhances feature extraction by adjusting lighting, texture, and noise levels, thereby reducing discrepancies between AI-rendered and real images. The results indicate that AI approaches, CNNs, hold immense potential for further improving classification robustness. This study contributes to the growing field of AI-driven interior design classification by presenting a high-accuracy model capable of distinguishing between real and AI-generated environments with unprecedented precision.