CumInCAD is a Cumulative Index about publications in Computer Aided Architectural Design
supported by the sibling associations ACADIA, CAADRIA, eCAADe, SIGraDi, ASCAAD and CAAD futures

id ecaade2017_041
authors Fukuda, Tomohiro, Kuwamuro, Yasuyuki and Yabuki, Nobuyoshi
year 2017
title Optical Integrity of Diminished Reality Using Deep Learning
source Fioravanti, A, Cursi, S, Elahmar, S, Gargaro, S, Loffreda, G, Novembri, G, Trento, A (eds.), ShoCK! - Sharing Computational Knowledge! - Proceedings of the 35th eCAADe Conference - Volume 1, Sapienza University of Rome, Rome, Italy, 20-22 September 2017, pp. 241-250
summary A new method is proposed to improve diminished reality (DR) simulations to allow the demolition and removal of entire buildings in large-scale spaces. Our research goal was to obtain optical integrity by using a scientific and reliable simulation approach. Further, we tackled presumption of the texture of the background sky by applying deep learning. Our approach extracted the background sky using information from the actual sky obtained from a photographed image. This method comprised two steps: (1) detection of the sky area from the image through image segmentation and (2) creation of an image of the sky through image inpainting. The deep convolutional neural networks developed by us to train and predict images were evaluated to be feasible and effective.
keywords Diminished Reality; Optical Integrity; Deep Learning; Augmented Reality; Landscape assessment
series eCAADe
email fukuda@see.eng.osaka-u.ac.jp
full text file.pdf (16,224,315 bytes)
references Content-type: text/plain
last changed 2017/09/13 13:21
HOMELOGIN (you are user _anon_868002 from group guest) Works Powered by SciX Open Publishing Services 1.002