Answers are grounded in this paper and cite their sources.
Reliable façade material information is needed for preliminary embodied carbon screening of existing buildings, yet such data are often unavailable where BIM models, as-built drawings, and detailed material records are missing. This study presents a pilot, screening-level, façade-focused Minimum Viable Data (MVD) workflow that links AI-derived visible façade material ratios to preliminary embodied carbon estimates while reporting uncertainty and operational limits. The workflow integrates Segment Anything Model 2 (SAM2) with a multimodal vision-language model to identify visible façade components and generate Façade Material Ratios (FMR), which represent the proportional coverage of exposed façade materials on the analyzed elevation. These outputs are linked to geospatial measurements and material carbon factors to examine how visual segmentation uncertainty propagates into material quantity and façade-level embodied carbon estimates. The workflow was evaluated on seven residential buildings in Atlanta, Georgia, under limited-data conditions including vegetation occlusion, shadowing, and irregular residential geometries. Statistical comparisons showed a positive association between AI-derived and manually annotated FMR values (R = 0.78), while a paired t-test did not detect a statistically significant mean difference within the evaluated sample (p = 0.186). These results indicate a positive association within the pilot sample, although they do not establish equivalence between the two annotation approaches. The results support façade-level material inventory screening and help identify cases that warrant follow-up audits in data-scarce settings, while remaining limited to the exposed façade layer rather than assembly-level LCA.
No takes yet. Share an insight, caveat, or question.
Hu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: