Defect Detection module (DEFINPECT) is a software module that process 3D models for detecting automatically defects as an alternative to on-site inspection.
The proposed software module executes a novel pipeline of processes that exploit the benefits of geometrical and deep learning based state of the art methods for identifying geometrical patches in the 3D model that is provided as input, that can be classified as defects and separate them from patches that are classified as geometric features.
DEFINPECT is a software model that automatically detects and highlights geometrical defects in 3D objects, separating them accurately from other geometrical patterns of their surface, which have similar geometrical meaning (e.g., high spatial frequency), but represent features of the object.
It can be used for remote inspection and tracking of defects in scanned 3D models facilitating tasks which traditionally required a considerable amount of human effort and specialized knowledge of experts.
DEFINSPECT introduces a novel pipeline that efficiently exploits the benefits of a geometrical and a deep learning approaches through a coarse-to-fine defect detection approach.
It includes an initial step for the data preparation and then a feature vectors estimation step for the initial/coarse defect detection. To further improve the results of our method, a CNN-based learning approach is used, which receives the output of the previous step, to fine-tune the detecting accuracy.
The proposed SW module relies on the observation that defects in cultural heritage objects mainly affect the model’s symmetry. The method provides a solution to the challenge of the identification of geometrical defects and separating them from other geometrical patterns of the 3D surface, which have similar geometrical meaning (e.g., high spatial frequency), but represent features of the object.
To the best of our knowledge, this is the first time that a method using 3D information data as input tries to identify geometrical defects from 3D objects and separate them from other geometrical patterns of their surface, which have similar geometrical meaning (e.g., high spatial frequency), but represent features of the object. The main contribution can be summarized below:
Yes. DEFINSPECT receives as input model information as an OBJ or JSON file and generates a face decision vector with length equal to the number of model faces and a vertex decision vector with length equal to the number of vertices.
In a “standalone” version it is a separate module in the form of a .dll that provides functionallities for other .dlls. In a cloud based version, DEFINSPECT is executed in gpu-oriented keras python API and communicates with other modules via a REST API driven by well established web frameworks (i.e. Flask).
Protection of endangered heritage objects or build- ings is highly challenging and usually requires on-site visual inspection of experts to detect if any further intervention of a professional is required.
Nevertheless, this type of inspection is time-consuming and not suitable for quantitative analysis. In this direction, the proposed SW module is developed for automatic defect detection that could be used to facilitate challenging tasks which traditionally required a considerable amount of human effort and specialized knowledge of experts (e.g., a detailed inspection of defects in a historical object).
Of course its requires as input the 3D model that have been generated from the scanning process.
All this type of tools were created mainly for help the cultural heritage practitioners, conservators, restorers, and curators, with no technical skills 3D imaging and graphics.
The developed ontology is primarily of interest to professionals in the field of monument restoration and assessment.
The models are addressed to all the professionals and companies that operate in the field of facility management: these subjects are entrusted with the programming and deciding all maintenance actions over the course of the life of a building by its owner, with the task of achieving the best results with the minimum expenditure. The model can be adopted by them for the management of a historical building, hence easing and improving the process of choosing the best maintenance plans and strategies for them.
First of all, we would like to thank you for your availability to collaborate in a study designed by researchers from the WARMEST´S project.
Your participation will be essential for the success of this work. Specifically, we would like you to answer honestly a series of questions about your experience with your volunteer tourism program. The estimated duration of the survey is about 10 minutes.
Remember that there are no correct answers and that your sincerity is essential to guarantee the rigor of the investigation.
The information you voluntarily provide and the data collected is completely anonymous and will be analyzed in an aggregated manner to carry out statistical analysis for scientific research purposes.
Kind regards and thanks again!
This survey is anonymous.
The record of your survey responses does not contain any identifying information about you, unless a specific survey question explicitly asked for it.
If you used an identifying token to access this survey, please rest assured that this token will not be stored together with your responses. It is managed in a separate database and will only be updated to indicate whether you did (or did not) complete this survey. There is no way of matching identification tokens with survey responses.