Electronic Resource
Automatic Damage Detection on Traditional Wooden Structures with Deep Learning-Based Image Classification Method
| العنوان: | Automatic Damage Detection on Traditional Wooden Structures with Deep Learning-Based Image Classification Method |
|---|---|
| Additional Titles: | Automatsko otkrivanje oštećenja na tradicionalnim drvenim konstrukcijama metodom klasifikacije slika utemeljenom na dubokom učenju |
| المؤلفون: | Haciefendioglu, Kemal, Başaga, Hasan Basri, Kartal, Murat Emre, Bulut, Mehmet Ceyhun |
| المصدر: | Drvna industrija; ISSN 0012-6772 (Print); ISSN 1847-1153 (Online); Volume 73; Issue 2 |
| بيانات النشر: | University of Zagreb Faculty of Forestry and Wood Technology 2022 |
| نوع الوثيقة: | Electronic Resource |
| مستخلص: | Wood has a long history of being used as a valuable resource when it comes to building materials. Due to various external factors, in particular the weather, wood is liable to progressive damage over time, which negatively impacts the endurance of wooden structures. Damage assessment is key in understanding, as well as in effectively mitigating, problems that wooden structures are likely to face. The use of a classification system, via deep learning, can potentially reduce the probability of damage in engineering projects reliant on wood. The present study employed a transfer learning technique, to achieve greater accuracy, and instead of training a model from scratch, to determine the likelihood of risks to wooden structures prior to project commencement. Pretrained MobileNet_V2, Inception_V3, and ResNet_V2_50 models were used to customize and initialize weights. A separate set of images, not shown to the trained model, was used to examine the robustness of the models. The three models were compared in their abilities to assess the possibilities and types of damage. Results revealed that all three models achieve performance rates of similar reliability. However, when considering the loss ratios in regard to efficiency, it became apparent that the multi-layered MobileNet_V2 classifier stood out as the most effective of the pre-trained deep convolutional neural network (CNN) models. Drvo kao vrijedan građevni materijal ima dugu povijest uporabe u graditeljstvu. No zbog brojnih vanjskih čimbenika, posebice vremenskih utjecaja, drvo tijekom vremena postaje podložno progresivnom propadanju, što negativno utječe na izdržljivost drvenih konstrukcija. Procjena šteta na drvu ključna je za razumijevanje problema koji će vjerojatno nastati na drvenim konstrukcijama, kao i za njihovo učinkovito ublažavanje. Primjena sustava klasifikacije uz pomoć dubokog učenja može potencijalno smanjiti vjerojatnost oštećenja u inženjerskim projektima koji se oslanjaju na drvo. U ovom je istraživanju primijenjena tehnika transfernog učenja kako bi se postigla veća točnost modela umjesto da se model za utvrđivanje vjerojatnosti rizika za drvene konstrukcije radi prije početka projekta. Za prilagodbu i inicijalizaciju težina primijenjeni su unaprijed osposobljeni modeli MobileNet_V2, Inception_V3 i ResNet_V2_50. Za ispitivanje robusnosti modela upotrijebljen je zaseban skup slika koji nije prikazan u osposobljenome modelu. Spomenuta tri modela uspoređena su s obzirom na njihove mogućnosti procjene vjerojatnosti i vrste oštećenja drvenih konstrukcija. Rezultati su otkrili da sva tri modela imaju sličnu pouzdanost. Međutim, kada se uzmu u obzir omjeri gubitaka u odnosu prema učinkovitosti, postalo je očito da se višeslojni MobileNet_V2 klasifikator istaknuo kao najučinkovitiji od unaprijed pripremljenih modela dubokih konvolucijskih neuronskih mreža (CNN). |
| مصطلحات الفهرس: | deep learning method; convolutional neural networks; MobileNet_V2; Inception_V3; ResNet_ V2_50; wooden structures, metoda dubokog učenja; konvolucijska neuronska mreža; MobileNet_V2; Inception_V3; ResNet_V2_50; drvene konstrukcije, text, info:eu-repo/semantics/article, info:eu-repo/semantics/publishedVersion |
| URL: | info:eu-repo/semantics/altIdentifier/doi/10.5552/drvind.2022.2108 |
| الاتاحة: | Open access content. Open access content info:eu-repo/semantics/openAccess Journal Drvna industrija is at the highest possible level of Open Access, meaning that all content is immediately and freely available to anyone, anywhere, to be downloaded, printed, distributed, read, reused, self archived, and re-mixed (including commercially) without restriction, as long as the author and the original source are properly attributed according to the Creative Commons Attribution 4.0 International License (CC BY). The author(s) hold the copyright and retain publishing rights without restrictions. CC BY (Creative Commons Attribution) is the most accommodating of public copyright licenses as defined by Creative Commons, a nonprofit organization that provides legal tools for sharing and use of creative works and research. The CC BY license is recommended for maximum dissemination and use of licensed materials. All content published in Journal of Drvna industrija is available under CC BY, meaning anyone is free to use and reuse the content provided the original source and authors are credited. The copyright is held and retained. The author(s) hold the copyright without restrictions. |
| ملاحظة: | application/pdf English |
| Other Numbers: | HRCAK oai:hrcak.srce.hr:278445 1349086539 |
| المصدر المساهم: | HRCAK PORTAL ZNANSTVENIH CASOPISA REPUB From OAIster®, provided by the OCLC Cooperative. |
| رقم الانضمام: | edsoai.on1349086539 |
| قاعدة البيانات: | OAIster |
| الوصف غير متاح. |