Search papers, labs, and topics across Lattice.
This study reformulates image-based defect detection for infrastructure inspection as image difference classification (IDC) to mitigate the challenges posed by limited annotated data. By applying this approach to low-resource traffic sign inspection, the researchers evaluated various IDC classifiers and found that an instruction-based classifier significantly outperformed encoder-based methods. The results highlight IDC's potential as a robust modeling strategy for enhancing asset condition monitoring in digital twins, particularly in data-scarce environments.
Instruction-based classifiers can dramatically improve defect detection in infrastructure inspection, outperforming traditional methods even in low-data scenarios.
Digital twins (DTs) allow the digitalization of road infrastructure inspection, though this is hindered by limited annotated data. This work exploits the relational nature of continuous asset condition monitoring to reformulate image-based defect detection as image difference classification (IDC) to reduce data reliance. This was evaluated in a case study on low-resource traffic sign inspection with different IDC classifiers using a newly-curated, high quality dataset. Results indicate that the instruction-based classifier outperforms encoder-based ones and gains from comparison with reference images. This shows that IDC can be an effective task modeling for tackling data constraints in infrastructure inspection and DT asset condition updating.
