[dipl] Perustieteiden korkeakoulu / SCI

Permanent URI for this collectionhttps://aaltodoc.aalto.fi/handle/123456789/21

Browse

Browsing [dipl] Perustieteiden korkeakoulu / SCI by Subject "2D images"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Similarity based search of 3D assemblies in Tekla Structures using 2D images
    (2022-07-29) Nybergh, Peter
     Perustieteiden korkeakoulu | Master's thesis
    Within the building information modeling (BIM) framework, and within the Tekla Structures (TS) software more specifically, there has been a growing need for finding assemblies based on structural similarity. Previous methods developed in TS to address this problem make use of assembly properties, which do not guarantee structural similarity. In general, the subject of assessing the similarity of 3D objects for object retrieval is a well-established field of research, particularly within 3D CAD modeling, but increasingly within the BIM framework as well. Image-based techniques, in particular, have proven increasingly successful in 3D object classification and retrieval in general. In this work, an image-based technique was investigated for the purpose of assembly retrieval using three orthogonal views of the wireframe of the assembly. To perform feature extraction, two different machine learning architectures were explored, namely deep autoencoder based models and models which utilized the deep neural network, Xception. The models were also integrated into TS, expanding on the previous property based methods and UI. Using assemblies extracted from BIM models of different constructions, two test cases were used to evaluate model performance. The findings show that the models are able to distinguish between assemblies, showing that there is potential for image-based approaches for assembly search methods within TS.