| Title |
Auto-recognition of Interior Design Images for Managing Architectural Design References - Focused on the Module Implementation for Recognizing the Usage of Rooms of Korean Apartments - |
| Authors |
김진성(Kim, Jin-Sung) ; 이진국(Lee, Jin-Kook) |
| DOI |
http://dx.doi.org/10.14774/JKIID.2018.27.5.013 |
| Keywords |
Interior Design Reference Images ; Deep Learning ; Image Automatic Recognition ; Intelligent Reference Management |
| Abstract |
This paper aims to suggest the approach to auto-recognition of the interior design reference image and its prototype application for intelligently managing the design cases. The reference images are significantly utilized in design process so that efficient management method is needed. The current approaches, however, rely on each of architects or designer, so it takes lots of time and human resource to collect, classify, save and search. Deep learning-based image recognition technique enable computer to understand the content of image. Using this technique and pre-trained model, we train the computer to recognize the Korean apartment room design image and infer its room usage. The 4500 images labeled with bathroom, bedroom, entrance, kitchen and living room are used for training. After training the model, testing is carried out with 1000 images that is not used on training. Analysis on how well the model recognize the image`s category is conducted with the correct answer rate and accuracy. The auto-recognition prototype application and its testing result show the possibility of the core modules for intelligent architectural design case management system and contributed to supporting for architect or designer to concentrating the design core work by reducing the non-design works such as reference data management. |