Innovative deep learning methods for improving qlassic assessment outcomes
As societal demands shift toward higher-quality living environments, construction defects have become a significant concern for homeowners. To address this, the Construction Industry Development Board Malaysia (CIDB) introduced the Quality Assessment System in Construction (QLASSIC), a system design...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Construction Research Institute of Malaysia
2025
|
| Subjects: | |
| Online Access: | https://umpir.ump.edu.my/id/eprint/47361/1/Innovative%20deep%20learning%20methods%20for%20improving%20qlassic%20assessment.pdf https://umpir.ump.edu.my/id/eprint/47361/ https://www.cream.my/prod/mcrj-volume-47-no-3-2025 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | As societal demands shift toward higher-quality living environments, construction defects have become a significant concern for homeowners. To address this, the Construction Industry Development Board Malaysia (CIDB) introduced the Quality Assessment System in Construction (QLASSIC), a system designed to evaluate construction quality based on defect quantification. However, traditional manual inspections are labour-intensive, costly, time-consuming, and subjective. Recent advancements in machine vision, particularly deep learning, have significantly improved automated defect detection. Convolutional Neural Networks (CNNs) are especially favoured for their ability to autonomously identify key features without human intervention. While numerous studies have explored the application of deep learning for defect detection, there remains a lack of integration with comprehensive systems such as QLASSIC. This paper presents an innovative approach that combines deep learning with the QLASSIC standard to assess housing quality. The proposed approach includes the following key steps: (1) categorisation of visual defects according to QLASSIC; (2) collection of defect-related image datasets; (3) image preprocessing and manual pixel-level annotation; (4) selection and training of deep learning models; and (5) evaluation and refinement of model performance. With the proposed approach, it is expected to improve the efficiency and accuracy of building assessments and subsequently foster clearer communication among all parties involved in the construction process. |
|---|