Pith. sign in

REVIEW 3 major objections 4 minor 47 references

The paper claims that automated building-level housing inspection works primarily by aggregating detections across multiple photographs of each building, with neighborhood POI context adding only a small, category-dependent correction.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 10:18 UTC pith:3CWLIZHD

load-bearing objection Multi-view aggregation is the real story; the POI gain is modest and likely inflated by selection on test data — fix the leakage and the paper holds up. the 3 major comments →

arxiv 2607.20263 v1 pith:3CWLIZHD submitted 2026-07-22 cs.CV

How Does Urban Context Relate to Residential Building Health? A Vision-POI Fusion Framework for Building-Level Housing Inspection

Kun Zhao , Helei Ren , Guilin Tang , Tianyi Chen , Zhehui Song , Xing Liu , Lijian Zhou , Yuhong Zhao
show 3 more authors
Xiang Gao Jinming Jiang Qichao Ban
This is my paper
classification cs.CV
keywords urban physical examinationresidential building healthobject detectionmulti-view aggregationpoints of interestbuilding-level inspectionspatial cross-validationvisual-GIS fusion
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Automated housing inspection at the building level can be substantially improved by aggregating object-detection outputs across the many photographs taken of each building, rather than by judging single images. The paper tests whether neighborhood context—the mix of shops, offices, and services around a community, derived from points-of-interest (POI) data—adds further predictive information. On a dataset of 92 old residential communities, 3,237 buildings, and 25,608 field images in Qingdao, multi-view aggregation raises building-level Macro-F1 from 60.84% to 74.95%. Adding POI context raises it to 76.79%, a modest, category-dependent gain; context alone reaches only 47.88%, so it cannot stand in for visual evidence. The practical claim is that POI context acts as a supplementary prior for prioritizing and interpreting inspection results, not as a causal or standalone signal.

Core claim

The discovery is that the main performance leap in building-level housing inspection comes from a simple aggregation step: take all field images linked to one building, run an object detector on each, and summarize the detections into a seven-component visual signature (whether anything was found, how many boxes, confidence statistics, and the fraction of views that contain the issue). This multi-view visual aggregation lifts Macro-F1 from 60.84% for per-image thresholding to 74.95%. Adding POI-derived neighborhood features through a cost-sensitive classifier under community-isolated spatial cross-validation yields 76.79%, an absolute gain of only +1.84 points over visual aggregation alone a

What carries the argument

The load-bearing object is the three-stage pipeline: (1) image-level object detection on each field photo; (2) building-level aggregation that converts per-image detections into a binary visual-existence flag plus six statistics (count, max/mean/median confidence, confidence standard deviation, and view-level occurrence rate); and (3) a post-correction classifier that fuses those visual features with POI-derived neighborhood features computed at 500m, 1000m, and 1500m buffers. The paper's main comparison isolates the contribution of each stage: Direct Detection (threshold any image), Multi-View Visual Aggregation (the aggregated visual features alone), Urban Functional Context Only (POI feat

Load-bearing premise

The load-bearing assumption is that the POI feature selection—correlation screening across the 92 communities—does not use information from communities that end up in the held-out test folds of the spatial cross-validation; the paper does not nest feature selection inside folds, so the reported +1.84 point POI gain may be inflated.

What would settle it

Re-run the full pipeline with POI feature selection (correlation screening and radius choice) performed inside each training fold only, then compare MSCF and MVVA under the same community-isolated folds. If the macro-F1 difference falls to zero or below, the claim that POI context adds predictive value fails for this dataset; a positive difference that survives nested selection would confirm it.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • A detector that wins on image-level mAP need not be the best front-end for building-level diagnosis; the paper finds the best image-level detector was not the best building-level front-end.
  • The 1000m POI buffer outperforms both 500m and 1500m, so neighborhood context has a scale sweet spot; the 1500m scale, despite the strongest univariate correlations, produces worse out-of-sample predictions.
  • Urban context alone is insufficient: POI-only input yields Macro-F1 of 47.88%, so POI should be used only as a supplementary feature on top of visual evidence.
  • Spatial leakage inflates results: ordinary cross-validation gives 81.50% Macro-F1 versus 76.79% under community-isolated cross-validation, so community-isolated evaluation is necessary for honest estimates.
  • For the sparsest categories, POI fusion improves screening precision but not reliable detection; pipeline damage reaches 21.43% F1 yet only 3 of 23 positive buildings are found.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A testable extension: nest the POI feature-selection step inside each training fold of the spatial cross-validation. If the MSCF-minus-MVVA F1 gain disappears, the reported +1.84 points is partly a selection-leakage artifact; if it persists, the contextual prior is real.
  • The paper's emphasis on F1 understates what a probabilistic post-corrector could deliver; POI context may add more value through probability ranking (ROC/PR-AUC) than through binary label changes, and future work should report risk-ranking metrics separately.
  • Because POI gains are small, the framework's transfer value likely lies in the multi-view aggregation step; a second city with different planning and POI ecosystems could be used to test whether any POI radius remains beneficial.
  • Static POI counts cannot capture transient activity; incorporating mobile-signaling or foot-traffic data, which the paper names as future work, would test whether dynamic context beats static context before investing in larger image datasets.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a three-stage vision–POI fusion framework for building-level housing inspection in old residential communities, using 3,237 buildings and 25,608 field images across 92 communities in Qingdao. Image-level object detection is performed with one of several detectors (YOLOv8, YOLOv9c, YOLOv10s, YOLO11n, RT-DETR-L), and detections are aggregated into six building-level visual features plus a binary visual-existence indicator. POI-derived neighborhood context is constructed at 500 m, 1,000 m, and 1,500 m buffers, with Pearson/Spearman correlation screening under FDR correction. A cost-sensitive Random Forest fuses visual and POI features under community-isolated Spatial Group CV. The main reported results are: Multi-View Visual Aggregation raises building-level Macro-F1 from 60.84% (Direct Detection) to 74.95%; adding POI context (MSCF) raises Macro-F1 to 76.79%, a gain of +1.84 points. The paper interprets POI context as a modest, category-dependent, non-causal supplementary prior.

Significance. If the central results hold, the paper makes a useful applied contribution: it demonstrates that building-level inspection can be improved substantially by aggregating multi-view image detections, and it provides a careful, community-isolated evaluation protocol for the fusion classifier. The paper is also transparent in several respects: it releases code and data links, reports class-level confusion matrices for sparse categories (Table 16), and explicitly frames POI features as predictive context rather than causal determinants. The class-wise diagnostic tables and the honest reporting of the limited POI gain are strengths. However, the second quantitative claim (the +1.84 pp POI gain) rests on a feature-selection and model-selection protocol that is not fully nested within the spatial cross-validation folds, and there is a potential overlap between the object detector's training communities and the test communities of the downstream classifier. These issues affect the headline MSCF-vs-MVVA comparison and require correction before the POI-context claim can be accepted.

major comments (3)
  1. [§3.3.2, §4.4, Table 14, Table 17] The POI contribution is quantified by comparing MSCF with MVVA under a configuration that is selected on the test folds. In §3.3.2, Pearson/Spearman correlations are computed over all 92 communities, including communities that later appear in test folds of the Spatial Group CV. In §4.4, the final setting 'YOLOv8 + 1000m' is chosen because it achieved the highest MSCF F1 on those same test folds (Table 14). The reported +1.84 pp gain is therefore the maximum over the POI feature set, buffer radius, and detector front-end, and Table 17 is explicitly descriptive. To support the claim that POI context provides an additional gain, the feature-selection step (including the FDR screening and the radius choice) must be nested inside each training fold, or the full configuration must be pre-registered before seeing test outcomes. Until then, the second quantitative claim is vulnerable to selectio
  2. [§2.4, §4.5, Table 10] The Spatial Group CV described in §4.5 is applied to the Random Forest classifier, but the visual features are produced by object detectors trained once on the whole image dataset under an 8:2 split (§2.4, Table 10). If that 8:2 split was not performed at the same community level as the 92-community folds, or if detector training communities overlap RF test communities, then the MVVA/MSCF test features are not genuinely out-of-sample. The paper should clarify whether the detector training/validation split was spatially isolated, and if not, it should retrain the detector within each spatial fold or demonstrate that this overlap has a negligible effect. This concern affects both the MVVA and MSCF results, including the central multi-view aggregation claim.
  3. [§3.3.2, §3.4, Table 6, Table 15] There is a mismatch between the POI screening results and the POI features used in MSCF. Table 6 reports that at the 1,000 m radius only 'Addition balconies windows' has significant POI correlations after FDR correction, yet Table 15 shows MSCF gains for Pipeline damage, Illegal renovation expansion, and Elevator addition under the main YOLOv8 + 1000m setting. The paper should state exactly which POI variables were used for each of the seven categories in the 1,000 m model. If variables that did not pass the FDR screening were nevertheless included, the screening criterion is not consistently applied; if no POI variables were included for those categories, the reported gains need a different explanation.
minor comments (4)
  1. [§3.3.2, §4.4] The references 'Figure 4-3', 'Figure 4-4', and 'Table 4-6' appear to refer to Figure 7 and Table 14. These cross-reference labels should be updated to the actual figure and table numbers.
  2. [§4.4, §4.6.2] The phrase 'spatial plausibility test' is used for the correlation analysis; consider renaming it to 'association screening' to avoid implying that a correlation is a test of spatial plausibility.
  3. [Table 17] The table reports descriptive paired differences without confidence intervals or significance tests. The accompanying caveat is welcome, but adding the standard deviation of the per-fold differences (or a paired bootstrap interval) would make the limited POI gain easier to interpret.
  4. [§2.4, §4.1] The 8:2 'spatial heterogeneity ratio' for the detector training/validation split is described only briefly. Please specify whether the split was by community, building, or image, because this directly affects the interpretation of the downstream spatial CV results.

Circularity Check

2 steps flagged

POI gain is partly test-selected (feature screening and radius/detector choice see test-fold labels); the central multi-view visual aggregation result is independent and robust.

specific steps
  1. fitted input called prediction [§3.3.2 Multi-scale Correlation Analysis (Equation 3; Tables 5-6)]
    "Where 𝑛 is the total number of sample communities (𝑛=92); 𝑥𝑖 refers to the proportion of a specific POI category in the 𝑖-th community; 𝑦𝑖 refers to the problem incidence rate under a specific health check-up problem in the 𝑖-th community ... This step aims to identify candidate environmental variables with statistical significance, rather than directly determining the final POI radius; the best-performing scale for final modeling still needs to be verified through the building-level Multi-Source Context Fusion (MSCF) experiments in Section 4.4."

    The FDR screening is computed on the target incidence rates of all 92 communities, i.e. on the same labels that define the Spatial Group CV test folds. The selected POI variables are then fed as x_e into MSCF, and the resulting out-of-sample F1 is reported as a predictive gain. Because test labels chose which POI inputs enter the classifier, the MSCF comparison is not a clean held-out prediction; the POI feature set is a fitted input whose reported contribution is partly informed by the test set.

  2. fitted input called prediction [§4.4 Multi-Scale POI Fusion Post-Correction Experiments (Table 14)]
    "the post-correction judgment performance of different visual object detection front-ends under 500m, 1000m, and 1500m radii was tested ... reached the best-performing value at the 1000m block scale in this dataset ... the decision combination using YOLOv8 as the front-end detector fused with a 1000m POI context (YOLOv8 + 1000m MSCF) recorded the highest overall building-level judgment score, with an average F1-score reaching 76.79% ... Therefore, this study determined YOLOv8 + 1000m as the main setting."

    The headline 76.79% is the maximum over the tested detector×radius grid and is computed on the Spatial Group CV test folds; that same test-fold result is then used to select YOLOv8 + 1000m as the main setting. The +1.84-point MSCF-vs-MVVA gain is therefore a test-selected best case rather than the out-of-sample performance of a pre-specified model. Table 17 calls the paired differences 'descriptive ... rather than formal significance tests,' but the selection bias from choosing the best configuration on the test set is not accounted for.

full rationale

The paper's primary quantitative claim—that multi-view aggregation lifts building-level Macro-F1 from 60.84% (DD) to 74.95% (MVVA)—is not circular: it compares raw image-triggered decisions with an RF classifier trained on multi-view visual features under community-isolated Spatial Group CV, and it does not depend on POI feature selection. I find no load-bearing self-citation: HOUSED is a publicly released dataset used as auxiliary training data, and the Qingdao field labels provide independent evaluation. The secondary claim (MSCF 76.79% vs MVVA 74.95%, +1.84) is partially circular in the weaker, information-leakage sense. §3.3.2 screens POI variables using Pearson/Spearman correlations over all 92 communities, including communities that later fall into test folds, and §4.4 selects the POI radius and detector by the highest MSCF F1 on the same test folds. Both steps cause test labels to influence the reported POI gain. The paper hedges by calling Table 17 descriptive and §4.4 a spatial plausibility test rather than proof of predictive superiority, and §5.3 candidly limits the study to within-city generalization, but these hedges do not remove the leakage. I also note an internal inconsistency (Table 6 shows only 'Addition balconies windows' significant at 1000m, while Table 15 reports MSCF gains for Pipeline damage, Illegal renovation expansion and Elevator addition at 1000m), but I do not treat that as circularity. Overall: central result independent; POI increment partly test-selected, so score 4, not 6+.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

No new physical or conceptual entities are introduced. The framework combines existing detectors, statistical tests, and classifiers. The main assumptions are about data quality and the validity of POI proximity as a proxy for context.

free parameters (5)
  • DD confidence threshold = 0.15
    Threshold for Direct Detection baseline and image-level evidence matching; chosen manually, not optimized.
  • POI buffer radii = 500/1000/1500 m
    Three hand-selected spatial scales approximating the 15-minute life circle; the final choice (1000 m) is selected based on test-set F1, a data-driven selection.
  • FDR significance threshold = pFDR < 0.05
    Standard threshold, but the number of selected features depends on it.
  • Random Forest decision threshold = not specified (default 0.5 assumed)
    The paper defines a decision threshold θ but does not state how it is set for each category.
  • Data balancing retention rates = 100% rare / 50% medium / 10% majority / 5% background
    Hand-chosen resampling probabilities in the two-stage balancing strategy (Section 2.4).
axioms (4)
  • domain assumption Field inspection records and their category mapping (15→7) are accurate ground truth for building-level issues.
    The supervised classifiers and all F1 scores depend on the correctness and consistency of the original municipal inspection labels (Sections 2.2, 4.5).
  • domain assumption POI counts within Euclidean buffers are a valid proxy for neighborhood functional context.
    The paper explicitly acknowledges road networks, barriers, and entrances are not modeled; it uses circular buffers around a community centroid (Section 2.3).
  • domain assumption The 6 aggregated visual features (count, confidence stats, view occurrence) are sufficient to characterize building-level visual evidence.
    No analysis tests whether other aggregations (e.g., detector-specific scores, spatial layout) would change the result (Section 3.2.2).
  • domain assumption Community-isolated split prevents spatial leakage in the classifier evaluation.
    Used to justify the main results, but feature selection is performed before splitting, so the assumption is only partially satisfied (Sections 3.3.2, 4.5).

pith-pipeline@v1.3.0-alltime-deepseek · 28417 in / 11555 out tokens · 98564 ms · 2026-08-01T10:18:01.081957+00:00 · methodology

0 comments
read the original abstract

Housing-level urban physical examination is essential for identifying residential building problems and supporting targeted urban renewal. Existing automated inspection studies primarily rely on individual images and rarely examine whether surrounding urban functional context can provide supplementary information for building-level assessment. This study proposes a vision-POI fusion framework that combines multi-view visual inspection with POI-derived neighborhood context for residential building health assessment. The empirical dataset covers 92 old residential communities, 3,237 residential buildings, and 25,608 field-acquired inspection images in Qingdao, China, encompassing seven categories of housing-related issues. First, multiple object detection models are evaluated to extract issue locations, categories, and confidence scores from individual images. The image-level outputs are subsequently aggregated across multiple views to construct interpretable building-level representations. Second, POI features are extracted within 500m, 1,000m, and 1,500m neighborhood buffers to characterize surrounding functional environments. Pearson and Spearman correlation analyses, combined with false discovery rate correction, are used to identify candidate contextual features. Finally, visual and POI features are integrated using a cost-sensitive Random Forest classifier under community-isolated spatial cross-validation. The results show that multi-view aggregation provides the main performance improvement, increasing the building-level Macro-F1 from 60.84% under Direct Detection to 74.95%. Incorporating POI context further increases Macro-F1 to 76.79%, although the additional gain is modest and category-dependent. POI information therefore functions as a supplementary contextual prior rather than a substitute for direct visual evidence or a causal determinant of building condition.

Figures

Figures reproduced from arXiv: 2607.20263 by Guilin Tang, Helei Ren, Jinming Jiang, Kun Zhao, Lijian Zhou, Qichao Ban, Tianyi Chen, Xiang Gao, Xing Liu, Yuhong Zhao, Zhehui Song.

Figure 1
Figure 1. Figure 1: Spatial distribution of the 92 sampled residential communities in Qingdao. The main map shows the administrative boundary of Qingdao and the locations of 11 spatial clusters of old residential communities (A–K), highlighted in orange. detections are aggregated into interpretable building￾level visual features, while multi-scale POI indicators are introduced as supplementary contextual priors to support the… view at source ↗
Figure 2
Figure 2. Figure 2: Spatial distribution of the seven categories of housing-related issues across the 92 sampled communities in Qingdao. Color intensity represents the number of recorded issue instances in each community. administrative districts and 27 subdistricts in Qingdao, in￾cluding 3,237 residential buildings and 25,608 multi-view inspection images. Each building was associated with an average of 7.91 images, with a me… view at source ↗
Figure 3
Figure 3. Figure 3: Four-dimensional visual analysis of the finalized balanced training set: (a) displays the number of annotated instances for the 7 core categories; (b) overlays bounding boxes in a normalized space to show geometric forms and common aspect ratios; (c) shows the distribution of center coordinates within the image coordinate system; (d) reflects the relationship between the width and height of annotated bound… view at source ↗
Figure 4
Figure 4. Figure 4: Overall flow chart of the "perception-association-correction" three-stage vision–POI fusion framework for urban health check-ups. analyzing field-acquired inspection images, modern object detection models can identify visible problems such as wall damage, pipeline deterioration, unauthorized renovation, and unauthorized balcony or window additions. However, detection performance may be affected by complex … view at source ↗
Figure 5
Figure 5. Figure 5: Radar chart of YOLOv8 validation performance under different data balancing and augmentation strategies. augmentation strategies. The first factor is data distribu￾tion, including the original long-tail distribution dataset and the balanced resampling dataset; the second factor is the data augmentation strategy, including: (i) traditional single￾image augmentation method; (ii) strong augmentation com￾posit… view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of image-level mAP of mainstream visual detectors. enabling a systematic comparison of different architectural designs in complex urban inspection scenarios. The related detection precision and computational efficiency results are shown in [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Correlation heatmaps of multi-scale POI fusion post-correction experiments. results support the spatial relevance of POI-derived func￾tional context, but they do not by themselves establish stable building-level predictive gain. The added value of POI fea￾tures is therefore evaluated separately by comparing MVVA and MSCF under community-isolated cross-validation. Based on the above spatial environmental pr… view at source ↗
Figure 8
Figure 8. Figure 8: Average MSCF F1-score under different visual front￾ends and POI radii. 1000m POI scale, whereas YOLOv9c-640 reached only 73.85% under the same setting. This discrepancy indicates that the best image-level detector is not necessarily the best front-end for building-level diagnosis. Image-level mAP mainly evaluates box localization and classification in indi￾vidual images, whereas building-level diagnosis de… view at source ↗
Figure 9
Figure 9. Figure 9: Average F1 of four ablation schemes under yolov8_balanced + 1000m condition. Because several issue categories are highly sparse, over￾all Accuracy alone can be misleading. In such cases, a clas￾sifier may obtain a very high Accuracy mainly by correctly identifying the large number of negative buildings, while still missing most positive cases [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 11
Figure 11. Figure 11: Comparison of ROC and PR curves for major fusion models under multi-class decision-making [PITH_FULL_IMAGE:figures/full_fig_p022_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Class-level F1 under administrative-scale spatial extrapolation stress tests: Subdistrict GroupCV vs. District GroupCV damage" directly dropped to zero, showing that extremely sparse samples cannot support cross-district predictions. 4.6.2. Descriptive and Algorithmic Validation of POI Auxiliary Effects Because the macro-average F1 improvement of MSCF relative to MVVA is limited, it is still necessary to … view at source ↗
Figure 13
Figure 13. Figure 13: Building-level spatial diagnostic maps for three types of stable health check-up issues. 76.54%, respectively, while GradientBoosting reached 74.71%. These results indicate that the multi-source feature fusion framework is not dependent on a single classifier family. This paper ultimately selected Random Forest as the main classifier, primarily valuing its practical advantages such as requiring no complex… view at source ↗
Figure 14
Figure 14. Figure 14: Elevator addition local spatial diagnostic triptych. 5. Discussion 5.1. Compensatory Mechanisms and Extrapolation Rigor of Multi-Scale Spatial Contexts The MSCF framework proposed in this study establishes an associative correction path between "microscopic physi￾cal disease perception" and "macroscopic urban functional context" at the methodological level. This study does not regard POIs as direct "physi… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

47 extracted references · 1 canonical work pages

  1. [1]

    2020 , author =

    Paths and strategies for sustainable urban renewal at the neighbourhood level: A framework for decision-making , journal =. 2020 , author =

    2020

  2. [2]

    2022 , issn =

    A novel methodology (WM-TCM) for urban health examination: A case study of Wuhan in China , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.ecolind.2022.108602 , author =

    arXiv 2022

  3. [3]

    75 of Jianke [2023] , year =

    Guiding Opinions on Comprehensively Carrying Out City Physical Examination Work (Document No. 75 of Jianke [2023] , year =

    2023

  4. [4]

    Standard for Planning and Design of Urban Residential Areas (

  5. [5]

    Standard for Residential Performance Evaluation (

  6. [6]

    2026 , issn =

    Augmenting urban planning with computer vision: A review of the state-of-the-art , journal =. 2026 , issn =. doi:https://doi.org/10.1016/j.scs.2026.107209 , author =

    arXiv 2026

  7. [7]

    Building Change Detection for Remote Sensing Images Using a Dual-Task Constrained Deep Siamese Convolutional Network Model , year =

    Liu, Yi and Pang, Chao and Zhan, Zongqian and Zhang, Xiaomeng and Yang, Xue , journal =. Building Change Detection for Remote Sensing Images Using a Dual-Task Constrained Deep Siamese Convolutional Network Model , year =

  8. [8]

    An Optimized Edge-Focused Siamese Network for Monitoring New Illegal Buildings Using Satellite Images , year =

    Du, Haode and Huang, Zhou and Zhang, Yi , journal =. An Optimized Edge-Focused Siamese Network for Monitoring New Illegal Buildings Using Satellite Images , year =

  9. [9]

    GU-Net3+: A Global-Local Feature Fusion Algorithm for Building Extraction in Remote Sensing Images , year =

    Liu, Yali and Ni, Cui and Wang, Peng and Yang, Dongqing and Yuan, Hexin and Ma, Chao , journal =. GU-Net3+: A Global-Local Feature Fusion Algorithm for Building Extraction in Remote Sensing Images , year =

  10. [10]

    2026 , issn =

    Towards smart city supervision: A detection pipeline for illegal buildings , journal =. 2026 , issn =. doi:https://doi.org/10.1016/j.engappai.2025.113052 , author =

    arXiv 2026

  11. [11]

    Perez, Husein and Tah, Joseph H. M. and Mosavi, Amir , title =. Sensors , volume =. 2019 , number =

    2019

  12. [12]

    2023 , doi =

    Automated joint 3D reconstruction and visual inspection for buildings using computer vision and transfer learning , journal =. 2023 , doi =

    2023

  13. [13]

    Buildings , VOLUME =

    Shin, Hyunkyu and Kim, Jonghoon and Kim, Kyonghoon and Lee, Sanghyo , TITLE =. Buildings , VOLUME =. 2023 , NUMBER =

    2023

  14. [14]

    Ensembles of Deep Neural Networks for the Automatic Detection of Building Facade Defects From Images , year =

    Interlando, Matteo and Pacifico, Maria Giovanna and Novellino, Antonio and Pastore, Vito Paolo , journal =. Ensembles of Deep Neural Networks for the Automatic Detection of Building Facade Defects From Images , year =

  15. [15]

    Sustainability , volume =

    Ho, Jongnam and Ahn, Yonghan and Shin, Hyunkyu , title =. Sustainability , volume =. 2025 , number =

    2025

  16. [16]

    Sensors , volume =

    Fan, Yue and Mai, Jinghua and Xue, Fei and Lau, Stephen Siu Yu and Jiang, San and Tao, Yiqi and Zhang, Xiaoxing and Tsang, Wing Chi , title =. Sensors , volume =. 2025 , number =

    2025

  17. [17]

    , booktitle =

    Han, Bingxin and Gao, Chuanxiang and Zhao, Zuoquan and Zhang, Jihan and Chen, Xi and Chen, Ben M. , booktitle =. A Comprehensive Framework for Automated Facade Defect Evaluation Using Deep Learning , year =

  18. [18]

    Buildings , volume =

    Lin, Xiaoxia and Meng, Yingzhou and Sun, Lin and Yang, Xiaodong and Leng, Chunwei and Li, Yan and Niu, Zhenyu and Gong, Weihao and Xiao, Xinyue , title =. Buildings , volume =. 2025 , number =

    2025

  19. [19]

    Sensors , volume =

    Liu, Daiming and Jin, Yongqiang and Yang, Yuan and Xiao, Zhenyang and Zhao, Zeming and Gao, Changling and Zhang, Dingcheng , title =. Sensors , volume =. 2026 , number =

    2026

  20. [20]

    Drone-Guided Cognitive Feature Aggregation for High-Rise Building Surface Defect Inspection , year =

    Meng, Baoxing and Wang, Chengqin , journal =. Drone-Guided Cognitive Feature Aggregation for High-Rise Building Surface Defect Inspection , year =

  21. [21]

    , author =

    Deep Learning-Based Crack Detection: A Survey: SD Nguyen et al. , author =. International Journal of Pavement Research and Technology , volume =. 2023 , publisher =

    2023

  22. [22]

    Buildings , volume =

    Chen, Yulong and Zhu, Zilong and Lin, Zhijie and Zhou, Youmei , title =. Buildings , volume =. 2023 , number =

    2023

  23. [23]

    Buildings , volume =

    Choi, Seung-Mo and Cha, Hee-Sung and Jiang, Shaohua , title =. Buildings , volume =. 2024 , number =

    2024

  24. [24]

    2025 , doi =

    Deep learning-enhanced smart ground robotic system for automated structural damage inspection and mapping , journal =. 2025 , doi =

    2025

  25. [25]

    Journal of Civil Structural Health Monitoring , volume=

    Building damage inspection method using UAV-based data acquisition and deep learning-based crack detection , author=. Journal of Civil Structural Health Monitoring , volume=. 2025 , publisher=

    2025

  26. [26]

    Structural Health Monitoring , volume =

    Udayanga GWKN Gamage and Xuanni Huo and Luca Zanatta and T Delbruck and Cesar Cadena and Matteo Fumagalli and Silvia Tolu , title =. Structural Health Monitoring , volume =. 2026 , doi =

    2026

  27. [27]

    and C.Bui, Doanh and Long, Duong Phi and Vo, Nguyen D

    Dinh, Minh and Bui, Vu L. and C.Bui, Doanh and Long, Duong Phi and Vo, Nguyen D. and Nguyen, Khang , booktitle =. Performance Evaluation of Optimizers for Deformable-DETR in Natural Disaster Damage Assessment , year =

  28. [28]

    International Journal of Disaster Risk Science , volume =

    A Deep Learning Application for Building Damage Assessment Using Ultra-High-Resolution Remote Sensing Imagery in Turkey Earthquake , author =. International Journal of Disaster Risk Science , volume =. 2023 , publisher =

    2023

  29. [29]

    Assessing the Generalization Ability of a Global Model for Rapid Building Damage Assessment in Real-world Disaster Scenarios , year =

    Edinç, Eren Berk and Bayazıt, Uluğ , booktitle =. Assessing the Generalization Ability of a Global Model for Rapid Building Damage Assessment in Real-world Disaster Scenarios , year =

  30. [30]

    Qiao, Wenfan and Shen, Li and Wang, Wei and Li, Zhilin , journal =. A Weakly Supervised Bitemporal Scene Change Detection Approach for Pixel-Level Building Damage Assessment Using Pre- and Post-Disaster High-Resolution Remote Sensing Images , year =

  31. [31]

    Remote Sensing , volume =

    Hatić, Damjan and Polushko, Vladyslav and Rauhut, Markus and Hagen, Hans , title =. Remote Sensing , volume =. 2025 , number =

    2025

  32. [32]

    2021 , author =

    Geo-registering UAV-captured close-range images to GIS-based spatial model for building façade inspections , journal =. 2021 , author =

    2021

  33. [33]

    Large-scale building dilapidation assessment for high-density cities: An urban visual intelligence approach , journal =

    Zihan Huang and Weisheng Lu and Junjie Chen and Yiyi Xie , doi =. Large-scale building dilapidation assessment for high-density cities: An urban visual intelligence approach , journal =. 2026 , issn =

    2026

  34. [34]

    2022 , issn =

    Mapping and modelling defect data from UAV captured images to BIM for building external wall inspection , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.autcon.2022.104284 , author =

    arXiv 2022

  35. [35]

    2024 , issn =

    Building defect inspection and data management using computer vision, augmented reality, and BIM technology , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.autcon.2024.105318 , author =

    arXiv 2024

  36. [36]

    2023 , doi =

    Damage-augmented digital twins towards the automated inspection of buildings , journal =. 2023 , doi =

    2023

  37. [37]

    Sensors , volume =

    Yang, Lei and Liu, Keju and Ou, Ruisi and Qian, Peng and Wu, Yunjie and Tian, Zhuang and Zhu, Changping and Feng, Sining and Yang, Fan , title =. Sensors , volume =. 2024 , number =

    2024

  38. [38]

    2025 , issn =

    Improvement of the defect inspection process of deteriorated buildings with scan to BIM and image-based automatic defect classification , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.jobe.2024.111601 , author =

    arXiv 2025

  39. [39]

    2014 , issn =

    Forecasting energy consumption of multi-family residential buildings using support vector regression: Investigating the impact of temporal and spatial monitoring granularity on performance accuracy , journal =. 2014 , issn =. doi:https://doi.org/10.1016/j.apenergy.2014.02.057 , author =

    work page doi:10.1016/j.apenergy.2014.02.057 2014

  40. [40]

    Environment and Planning B: Planning and Design , volume =

    Xingjian Liu and Ying Long , title =. Environment and Planning B: Planning and Design , volume =. 2016 , doi =

    2016

  41. [41]

    2023 , issn =

    Analyzing spatial patterns and influencing factors of different illegal land use types within ecological spaces: A case study of a fast-growing city , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.jclepro.2023.138883 , author =

    arXiv 2023

  42. [42]

    Remote Sensing , VOLUME =

    Muhmad Kamarulzaman, Aisyah Marliza and Wan Mohd Jaafar, Wan Shafrina and Mohd Said, Mohd Nizam and Saad, Siti Nor Maizah and Mohan, Midhun , TITLE =. Remote Sensing , VOLUME =. 2023 , NUMBER =

    2023

  43. [43]

    Sustainability , VOLUME =

    Sun, Haozun and Xu, Hong and He, Hao and Wei, Quanfeng and Yan, Yuelin and Chen, Zheng and Li, Xuanhe and Zheng, Jialun and Li, Tianyue , TITLE =. Sustainability , VOLUME =. 2023 , NUMBER =

    2023

  44. [44]

    ISPRS International Journal of Geo-Information , VOLUME =

    Wang, Jiwu and Hu, Yali and Duolihong, Wuxihong , TITLE =. ISPRS International Journal of Geo-Information , VOLUME =. 2023 , NUMBER =

    2023

  45. [45]

    2024 , doi =

    Investigating the effects of urban morphology on vitality of community life circles using machine learning and geospatial approaches , journal =. 2024 , doi =

    2024

  46. [46]

    ISPRS International Journal of Geo-Information , VOLUME =

    Guo, Zixuan and Xu, Hong and Lin, Qiushuang , TITLE =. ISPRS International Journal of Geo-Information , VOLUME =. 2025 , NUMBER =

    2025

  47. [47]

    2025 , issn =

    From points to patterns: An explorative POI network study on urban functional distribution , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.compenvurbsys.2024.102246 , author =

    arXiv 2025