Convolutional Neural Network with Pyramid Attention Mechanism for Floor Plan Layout Interior Design

Floor tiles are a most primary construction materials employed in floor decoration. The production process of floor tiles has adverse environmental effects. However, it is difficult to optimize space utilization while ensuring accessibility and functionality for diverse user needs. This research proposes a Convolutional Neural Network with Pyramid Attention Mechanism (CNN-PAM) for floor plan layout interior design. CNN-PAM enhances performance by focusing on appropriate spatial information at various scales which improves accuracy in recognizing complex floor plan layouts interior design. The CNN-NAM helps in capturing fine details such as room boundaries while managing spatial relationships across an entire layout. Initially, data is gathered from prior entries of the Solar Decathlon (SD) competition from 2007 to 2018. In the preprocessing step, Uniform annotation, Uniform drawing, Uniform labeling, and data augmentation are applied to form uniformity and enhance data. The proposed CNN-NAM achieves a high accuracy of 97.21% compared to existing methods like CNN-Non local Attention Mechanism (NAM), CNN-Cross-Scale NAM (CSNAM).