Data-driven interior plan generation for residential buildings

Abstract

We propose a novel data-driven technique for automatically and efficiently generating floor plans for residential buildings with given boundaries. Central to this method is a two-stage approach that imitates the human design process by locating rooms first and then walls while adapting to the input building boundary. Based on observations of the presence of the living room in almost all floor plans, our designed learning network begins with positioning a living room and continues by iteratively generating other rooms. Then, walls are first determined by an encoder-decoder network, and then they are refined to vector representations using dedicated rules. To effectively train our networks, we construct RPLAN - a manually collected large-scale densely annotated dataset of floor plans from real residential buildings. Intensive experiments, including formative user studies and comparisons, are conducted to illustrate the feasibility and efficacy of our proposed approach. By comparing the plausibility of different floor plans, we have observed that our method substantially outperforms existing methods, and in many cases our floor plans are comparable to human-created ones.

VR content creation and exploration with deep learning: A survey

“…Its auto-encoder performs scene object grouping while encoding information about objects' spatial properties, and scene generation during decoding which turns a randomly sampled code from the learned distribution into a plausible indoor scene hierarchy. Wu et al [94] proposed a data-driven method to automatically and efficiently generate floor plans for residential buildings given only the boundary. To do so, they created a large-scale dataset (RPLAN) consisting of real floor plans from residential buildings.…”

SCUT-AutoALP: A Diverse Benchmark Dataset for Automatic Architectural Layout Parsing

“…Previous studies regard Au-toALP as a pattern computing problem and many methods were introduced from the fields of computer vision, graphics or machine learning [8]- [10]. Most recently, the reviving of neural networks advance the studies by deep learning, which allows to build an end-to-end AutoALP system with trainable representation, generator [4]- [7] or predictor [1]- [3].…”

“…Many datasets have not been released publicly yet, which makes the comparison difficult. Recently, Wu et al [4] constructed a large-scale dataset containing over 80K real floor plans with dense layout anotations, and proposed a two-stage method with deep neural networks to generate floor plans for residential buildings with given boundaries. Kato et al [1] built a predictor for user preference of real estate properties based on a dataset containing 1.5K samples with 10 classes.…”

“…1 (b)) and layout label ( Fig. 1 (c)) allow interior plan generation as in [4]; samples ( Fig. 1 (a)) and layout labels ( Fig.…”

“…From a computational perspective, AutoALP has two subtasks. One is for prediction task, whose goal is to predict some architectural attributes [1]- [3]; the other is for generation task, which aims to generate candidate designs for the users [4]- [8], [10].…”