Artificial Intelligence-Assisted Interior Layout Design of CAD Painting

Abstract

As of today, with the promotion and application of artificial intelligence, a large amount of manual labor the current one, and its application to interior layout design will inevitably promote interior layout. Design innovation and optimization can also ensure the quality of modern interior layout design and effectively improve efficiency. This paper utilizes adversarial learning to design a conditional generative adversarial network (CGAN)-based approach for indoor scene layout estimation to predict the spatial layout structure of a room. Firstly, aiming at the problem that the boundary line of the layout edge map is easily blurred by the interpolation enlargement, a strategy of increasing the depth of the convolution layer and the deconvolution layer is adopted, and a new encoder-decoder network (Encoder-Decoder) is proposed to construct a conditional generative confrontation. A generative network for the network that produces a layout edge map of the same size as the original image. Then, for the difficult convergence problem of generative adversarial network training, a multi-scale strategy is used to build a multi-scale supervised network of generative networks to accelerate the convergence. The experimental results and analysis of the LSUN and Hedau standard datasets show that, compared with other layout estimation methods, this method can understand the indoor scene layout from an overall perspective and more accurately predict the 3D spatial layout structure of the room.