Intelligent Layout of Interior Design Using CenterNet Algorithm with 3D Modelling Interface

Nowadays, the interior space planning is an essential technique to effectively build an intelligent layout of interior design. This approach enabled the automatic generation of functional, efficient and visual appealed interior layouts. However, the traditional interior design methods heavily relied on manual planning and expertise. To overcome this issue, CenterNet algorithm with 3D Modelling Interface (CenterNet-3D modelling) is proposed. Initially, input data is collected from indoor space dataset and then preprocessed with horizontal flip, gaussian noise techniques. The horizontal flip maintains the structural integrity and gaussian noise adds random variations to expand input images. After that, the CenterNet algorithm is employed to detect objects in indoor images and further fed into Convolutional Neural Network with Gated Recurrent Unit (CNN-GRU) to capture long-range spatial relationships. Finally, the captured relationships are converted into vector formats to create 3D with transformation matrix for effective layout design. From the results, the proposed CenterNet-3D modelling attained better results in terms of accuracy (98.4%), precision (97.6%), recall (96.8%) and F1-score (97.5%) when compared to existing Deep Computer Aided Design (CAD).