Furnishing and rendering indoor scenes has been a longstanding task for interior design, where artists create a conceptual design for the space, build a 3D model of the space, decorates, and then performs rendering. Traditionally, the task is typically performed using professional 3D CAD design software, which is a tedious process and requires extensive prior knowledge and experience. Hence, we introduce the task of neural scene decoration (NSD), utilizing generative neural networks in assisting domain-specific scene synthesis.

Given a photograph of an empty indoor space and a list of decorations with layout determined by user, we aim to synthesize a new image of the same space with desired furnishing and decorations. Neural scene decoration can be applied to create conceptual interio...[ Read more ]

Furnishing and rendering indoor scenes has been a longstanding task for interior design, where artists create a conceptual design for the space, build a 3D model of the space, decorates, and then performs rendering. Traditionally, the task is typically performed using professional 3D CAD design software, which is a tedious process and requires extensive prior knowledge and experience. Hence, we introduce the task of neural scene decoration (NSD), utilizing generative neural networks in assisting domain-specific scene synthesis.

Given a photograph of an empty indoor space and a list of decorations with layout determined by user, we aim to synthesize a new image of the same space with desired furnishing and decorations. Neural scene decoration can be applied to create conceptual interior designs in a simple yet effective manner. In this work, a novel approach is proposed towards solving this problem, based on training a conditional GAN network. The performance of the proposed method is illustrated by comparing it with baselines built upon prevailing image translation approaches both qualitatively and quantitatively. In addition, extensive experiments are conducted to further validate the plausibility and aesthetics of the generated results based on the proposed approach.