Architectural Drawing Recognition A case study for training the learning algorithm with architectural plan and section drawing images

Architectural Drawing Recognition A case study for training the learning algorithm with architectural plan and section drawing images

Uzun, Can; Çolakoğlu, Meryem Birgül

Article:

This paper aims to develop a case study for training an algorithm to recognizearchitectural drawings. In order to succeed that, the algorithm is trained withlabeled pixel-based, architectural drawing (plan and section) dataset. During thetraining process, transfer learning (pre-training model) is applied. Thesupervised learning and convolutional neural network are utilized. After certainiterations, the algorithm builds awareness and can classify pixel-based plan andsection drawings. When the algorithm is shown a section that is not producedwith conventional drawing technic but through hybrid technics, it could predictthe drawing class correctly with %80 of accuracy. On the other hand, some of thealgorithm prediction is misoriented. We examined this prediction problem in thediscussion section. The results illustrate that neural networks are successful intraining algorithms to recognize and classify pixel-based architectural drawings.But for a highly accurate algorithm prediction, the dataset of the drawing imagesmust be ordered, according to sample resolution, sample size and samplecoherence for the dataset.

This paper aims to develop a case study for training an algorithm to recognizearchitectural drawings. In order to succeed that, the algorithm is trained withlabeled pixel-based, architectural drawing (plan and section) dataset. During thetraining process, transfer learning (pre-training model) is applied. Thesupervised learning and convolutional neural network are utilized. After certainiterations, the algorithm builds awareness and can classify pixel-based plan andsection drawings. When the algorithm is shown a section that is not producedwith conventional drawing technic but through hybrid technics, it could predictthe drawing class correctly with %80 of accuracy. On the other hand, some of thealgorithm prediction is misoriented. We examined this prediction problem in thediscussion section. The results illustrate that neural networks are successful intraining algorithms to recognize and classify pixel-based architectural drawings.But for a highly accurate algorithm prediction, the dataset of the drawing imagesmust be ordered, according to sample resolution, sample size and samplecoherence for the dataset.

Palavras-chave:

DOI: 10.5151/proceedings-ecaadesigradi2019_171

Referências bibliográficas

• [1]

Como citar:

Uzun, Can; Çolakoğlu, Meryem Birgül; "Architectural Drawing Recognition A case study for training the learning algorithm with architectural plan and section drawing images", p-29-34. In: Proceedings of 37 eCAADe and XXIII SIGraDi Joint Conference, “Architecture in the Age of the 4Th Industrial Revolution”, Porto 2019, Sousa, José Pedro; Henriques, Gonçalo Castro; Xavier, João Pedro (eds.). São Paulo: Blucher, 2019.
ISSN 23186968, DOI 10.5151/proceedings-ecaadesigradi2019_171