Research Output
Automatic Architectural Drawing Labelling Using Deep Convolutional Neural Network
  Architectural designers and technologists are able to make an assessment on buildability, thermal and hygrothermal performance of design details. To process drawings, human vision segments, classifies and distinguishes the drawing objects on the basis of their knowledge. With the rapid advancement of Artificial Intelligence methods, vast opportunities become available for performing tasks that used to require human intelligence or assistance by humans. Image processing and analysis is one of these tasks that consists of the manipulation of images using algorithms. There are various applications in different fields, and the use of it is increasing exponentially. This paper explores the use of image processing in identifying building materials in order to check compliance with building regulations and identify anomalies. In this paper, an encoder-decoder based deep convolutional neural network (DRU-net) for image segmentation is applied on architectural images to segment various materials including insulations, bricks and concrete in the conceptual development phase. An experimental analysis is performed on numerous detail drawings and an evaluation is made by mathematical models.


Sajjadian, S. M., Jafari, M., & Chen, X. (2021). Automatic Architectural Drawing Labelling Using Deep Convolutional Neural Network. In Sustainability in Energy and Buildings 2021 (69-78).



Image Processing, Architectural Detail Drawings, Deep Learning

Monthly Views:

Available Documents