Can Large Multimodal Models Inspect Buildings? A Hierarchical Benchmark for Structural Pathology Reasoning

TL;DR

This paper introduces a hierarchical benchmark called DefectBench to evaluate large multimodal models in building inspection tasks, highlighting their strengths in semantic understanding and topological reasoning but also their localization limitations.

Contribution

It presents a standardized, multi-dimensional benchmark and dataset for assessing LMMs in structural pathology reasoning, advancing evaluation standards in civil engineering AI applications.

Findings

LMMs excel in semantic perception and topological awareness.

LMMs show significant gaps in metric localization accuracy.

Zero-shot generative segmentation can match supervised models without domain training.

Abstract

Automated building facade inspection is a critical component of urban resilience and smart city maintenance. Traditionally, this field has relied on specialized discriminative models (e.g., YOLO, Mask R-CNN) that excel at pixel-level localization but are constrained to passive perception and worse generization without the visual understandng to interpret structural topology. Large Multimodal Models (LMMs) promise a paradigm shift toward active reasoning, yet their application in such high-stakes engineering domains lacks rigorous evaluation standards. To bridge this gap, we introduce a human-in-the-loop semi-automated annotation framework, leveraging expert-proposal verification to unify 12 fragmented datasets into a standardized, hierarchical ontology. Building on this foundation, we present \textit{DefectBench}, the first multi-dimensional benchmark designed to interrogate LMMs beyond basic semantic recognition. \textit{DefectBench} evaluates 18 state-of-the-art (SOTA) LMMs across three escalating cognitive dimensions: Semantic Perception, Spatial Localization, and Generative Geometry Segmentation. Extensive experiments reveal that while current LMMs demonstrate exceptional topological awareness and semantic understanding (effectively diagnosing "what" and "how"), they exhibit significant deficiencies in metric localization precision ("where"). Crucially, however, we validate the viability of zero-shot generative segmentation, showing that general-purpose foundation models can rival specialized supervised networks without domain-specific training. This work provides both a rigorous benchmarking standard and a high-quality open-source database, establishing a new baseline for the advancement of autonomous AI agents in civil engineering.