AquaMap-XAI: A Robustness-Aware and Explainable Transfer Learning Framework for Urban Water Issue Detection
DOI:
https://doi.org/10.4108/eetiot.13710Keywords:
Urban water issue detection, smart governance, transfer learning, MobileNetV3, robustness analysis, Grad-CAM, explainable AI, civic issue monitoringAbstract
Urban water-related problems such as drainage blockage, road flooding, water leakage, potholes, and visible contamination often require quick identification before civic authorities can take appropriate action. In many cities, citizens already capture and report such issues through mobile devices, but the submitted images still need to be verified and categorized before they can support decision-making. This work presents AquaMap-XAI, an explainable and robustness-aware deep learning framework for classifying urban water issue images. The work extends the preliminary AquaMap study by shifting from a single baseline classifier to a more systematic evaluation of transfer learning models, reliability through cross-validation, robustness under degraded image conditions, and visual explanation using Grad-CAM. Four pretrained convolutional models, namely ResNet18, MobileNetV3, EfficientNet-B0, and DenseNet121, were fine-tuned on a six-class AquaMap dataset. The models were evaluated using accuracy, precision, recall, Macro-F1 score, training time, and inference time. MobileNetV3 achieved the best overall performance with an accuracy of 0.88 and a Macro-F1 score of 0.87. Stratified five-fold cross-validation supported the stability of the selected model, while robustness analysis showed that Gaussian blur produced the highest performance degradation. Grad-CAM visualizations were used to inspect whether the model attended to meaningful regions related to water issues. The results indicate that AquaMap-XAI can provide a lightweight, interpretable, and practically useful direction for AI-assisted urban water issue monitoring and smart governance support.
Downloads
References
[1] Rammile S, Bhanye JI, Lehobo T, Hugo M. Building resilient urban infrastructure and innovation for urban public health: A case study of Bloemfontein, South Africa. In Sustainable development goals and urban health: Strides, challenges and way forward for poor neighborhoods 2024 Sep 29 (pp. 69-96). Cham: Springer Nature Switzerland.
[2] Sharma R, Katoch OR. Analysis of the targets and progress toward meeting the 2030 sustainable development goal-6 on clean water and sanitation: evidence from India. South Asian Journal of Social Studies and Economics. 2022 Sep 13;15(3):16-26.
[3] UN-HABITAT. World cities report 2024: Cities and climate action. Stylus Publishing, LLC; 2024 Nov 5.
[4] Parmesan C, Morecroft MD, Trisurat Y, Mezzi D. Climate change 2022: Impacts, adaptation and vulnerability (Doctoral dissertation, GIEC; IPCC).
[5] Wittemyer R, Bailur S, Anand N, Park KR, Gigler BS. New routes to governance: a review of cases in participation, transparency, and accountability. Closing the Feedback Loop. 2014:43.
[6] Wang L, Qian C, Kats P, Kontokosta C, Sobolevsky S. Structure of 311 service requests as a signature of urban location. PloS one. 2017 Oct 17;12(10):e0186314.
[7] Obuchowicz R, Strzelecki M, Piórkowski A. Clinical applications of artificial intelligence in medical imaging and image processing—A review. Cancers. 2024 May 14;16(10):1870.
[8] Yu F, Xiu X, Li Y. A survey on deep transfer learning and beyond. Mathematics. 2022 Oct 3;10(19):3619.
[9] Panda A, Panigrahi D, Mitra S, Mittal S, Rahimi S. Transfer learning applied to computer vision problems: survey on current Progress, limitations, and opportunities. arXiv preprint arXiv:2409.07736. 2024 Sep 12.
[10] He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition 2016 (pp. 770-778).
[11] Howard A, Sandler M, Chu G, Chen LC, Chen B, Tan M, Wang W, Zhu Y, Pang R, Vasudevan V, Le QV. Searching for mobilenetv3. In Proceedings of the IEEE/CVF international conference on computer vision 2019 (pp. 1314-1324).
[12] Tan M, Le Q. Efficientnet: Rethinking model scaling for convolutional neural networks. InInternational conference on machine learning 2019 May 24 (pp. 6105-6114). PMLR.
[13] Huang G, Liu Z, Van Der Maaten L, Weinberger KQ. Densely connected convolutional networks. In Proceedings of the IEEE conference on computer vision and pattern recognition 2017 (pp. 4700-4708).
[14] Nalla HL, Bhuchupalli A, Addala T, Sabbineni Y, Geetha KS, Aasha G, Bonthu S. AquaMap: Empowering Communities to Report and Map Water-Related Issues in Real-Time with Deep Learning. InInternational Conference on Cognitive Computing and Cyber Physical Systems 2024 Apr 4 (pp. 296-305). Cham: Springer Nature Switzerland.
[15] Van Calster B, Collins GS, Vickers AJ, Wynants L, Kerr KF, Barreñada L, Varoquaux G, Singh K, Moons KG, Hernandez-Boussard T, Timmerman D. Evaluation of performance measures in predictive artificial intelligence models to support medical decisions: overview and guidance. The Lancet Digital Health. 2025 Dec 13.
[16] Liu C, Dong Y, Xiang W, Yang X, Su H, Zhu J, Chen Y, He Y, Xue H, Zheng S. A comprehensive study on robustness of image classification models: Benchmarking and rethinking. International Journal of Computer Vision. 2025 Feb;133(2):567-89.
[17] Almulhim AI. Building urban resilience through smart city planning: a systematic literature review. Smart Cities. 2025 Jan 31;8(1):22.
[18] Asimakopoulos G, Antonopoulou H, Giotopoulos K, Halkiopoulos C. Impact of information and communication technologies on democratic processes and citizen participation. Societies. 2025 Feb 18;15(2):40.
[19] Chen LW, Liao CW, Liu JX. Proactive crowdsourced monitoring and sensing with expansible activity recognition based on Internet of Things Localization. IEEE Internet of Things Journal. 2025 Feb 5;12(11):17674-86.
[20] Yorucu V, Bekun FV, Yitmen I. The Anatomy of Smart Cities: A Global Perspective. In Modern Approach of Resilient and Sustainable Smart Cities 2026 Jan 24 (pp. 7-19). Cham: Springer Nature Switzerland.
[21] Hou Y, Chen X, Dong Q, Zhang F, Sun Y, Wu L, Liu Y. Enhancing drive-by sensing power in urban hotspots through multi-objective vehicle selection optimization. International Journal of Geographical Information Science. 2026 Jan 2;40(1):49-77.
[22] Sarika S, Hanan A, Afna VN, Johnson J, Nair KV. Agentic AI-Driven Grievance Redressal: Automating Categorization and Routing in Smart Governance. In2026 International Conference on Innovative Trends in Information Technology (ICITIIT) 2026 Mar 27 (pp. 1-6). IEEE.
[23] Darmawan MA, Mukti SN, Tahar KN. Pothole Detection Based on UAV Photogrammetry. Revue Internationale de Geomatique. 2025 Jan 1;34(1).
[24] Bhatt AK, Raj H, Sharma VB, Biswas S, Singh A, Silori R, Pandey M. Advancements in pothole detection techniques: a comprehensive review and comparative analysis. Discover Artificial Intelligence. 2025 Oct 3;5(1):255.
[25] Alam MI. Dual-Task Real-Time Low-Light Lane and Pothole Detection for Resource-Constrained Environments (Doctoral dissertation, IUB).
[26] Safyari Y. Pothole detection using machine learning and computer vision techniques (Doctoral dissertation, Memorial University of Newfoundland).
[27] Jayhind Pardeshi P, Shailja, Chaudhary A, Arya, Yadav M. Patch-Based Self-Supervised Learning for Road Damage Classification: A Case Study on the RDD2022 dataset of Indian test site. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences. 2025 Dec 19; 10:459-66.
[28] Awasthi YY, Chinzvende J. Deep learning-based flood forecasting using satellite imagery and IoT sensor fusion.
[29] Shi G, Chen Z, Lu R. Improved lightweight water body recognition model using Mobile-Former network. Hydrology Research. 2025 Sep 1;56(9):937-50.
[30] Zhao Z, Lu B, Zhang S, Wang D, Wan J, Liu R, Li H, Wang Z. Evaluating the performance of automatic detection for urban flood levels using different deep learning approaches. Water Resources Management. 2026 May;40(7):280.
[31] Gholizade M, Soltanizadeh H, Rahmanimanesh M, Sana SS. A review of recent advances and strategies in transfer learning. International Journal of System Assurance Engineering and Management. 2025 Mar;16(3):1123-62.
[32] Liu M, Wei J, Liu Y, Davis J. Human and ai perceptual differences in image classification errors. In Proceedings of the AAAI Conference on Artificial Intelligence 2025 Apr 11 (Vol. 39, No. 13, pp. 14318-14326).
[33] Khan D, Khan N, Ullah S. Harnessing hybrid intelligence and explainable AI for urban growth prediction: A Data-Driven framework for sustainable cities. Environment, Development and Sustainability. 2025 Sep 20:1-40.
[34] Zheng S, Cui X, Sun Y, Li J, Li H, Zhang Y, Chen P, Jing X, Ye Z, Yang L. Benchmarking pathclip for pathology image analysis. Journal of Imaging Informatics in Medicine. 2025 Feb;38(1):422-38.
[35] Xu B, Yang G. Interpretability research of deep learning: A literature survey. Information Fusion. 2025 Mar 1; 115:102721.
[36] Mothkur R, Soubhagyalakshmi P. Grad-CAM Based Visualization for Interpretable Lung Cancer Categorization Using Deep CNN Models. Journal of Electronics, Electromedical Engineering, and Medical Informatics. 2025 May 3;7(3):567-80.
[37] Andrae S. Explainable Artificial Intelligence. In Encyclopedia of Modern Artificial Intelligence 2026 (pp. 1-21). IGI Global Scientific Publishing.
[38] Mathew DE, Ebem DU, Ikegwu AC, Ukeoma PE, Dibiaezue NF. Recent emerging techniques in explainable artificial intelligence to enhance the interpretable and understanding of AI models for human. Neural Processing Letters. 2025 Feb 7;57(1):16.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Sridevi Bonthu, Padma Jyothi Uppalapati, Sankara Rao G
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
This is an open-access article distributed under the terms of the Creative Commons Attribution CC BY 4.0 license, which permits unlimited use, distribution, and reproduction in any medium so long as the original work is properly cited.
