Hybrid Neural Network Architecture with Adaptive Loss Function for Semantic Segmentation of Residential Building Defects
DOI:
https://doi.org/10.52575/2687-0932-2026-53-2-416-423Keywords:
semantic segmentation, cracks, HybridUNet, Vision Transformer, adaptive loss function, EfficientNet, spatial attentionAbstract
The article is focused on the problem of automating crack detection and segmentation in residential buildings using deep neural network architectures. A hybrid model named HybridUNet is proposed, combining an EfficientNet-B0 encoder pre-trained on ImageNet, a transformer bottleneck for modeling long-range spatial dependencies, and spatial attention modules in the decoder to suppress background noise. An adaptive loss function is applied, dynamically adjusting the contribution of Dice, Focal, and Boundary components depending on the epoch: the Dice loss weight is fixed at 1, the Focal loss weight linearly decreases from 1 to 0.5, while the Boundary loss weight increases from 0.1 to 1. This allows the model to first focus on the global defect structure and then refine boundaries. Comparison with baseline U‑Net models (with ResNet34 and EfficientNet‑B0 encoders) and DeepLabV3+ is performed on an open crack dataset. Experiments show that the proposed architecture achieves the best crack IoU (0.6029) and F1 score (0.7369) on the test set, outperforming DeepLabV3+ by 6.6% in IoU. The adaptive loss improves boundary accuracy, while the transformer block ensures robustness to noise and partial occlusions. The obtained results confirm the applicability of hybrid architectures with adaptive loss tuning for automated inspection of residential building stock.
Downloads
References
References
Abbasi S., Wahd A.S., Ghosh S., Ezzelarab M., Panicker M., Chen Y.T., Jaremko J.L., Hareendranathan A. 2025. Improved A-Line and B-Line Detection in Lung Ultrasound Using Deep Learning with Boundary-Aware Dice Loss. Bioengineering, 12(3): 311. DOI: 10.3390/bioengineering12030311.
Ahmed W., Liatsis P. 2025. Lhu-vt: A lightweight hypercomplex u-net with vessel thickness-guided dice loss for retinal vessel segmentation. Computers in Biology and Medicine, 185: 109470. DOI: 10.1016/j.compbiomed.2024.109470.
Azad R., Aghdam E.K., Rauland A., Jia Y., Avval A.H., Bozorgpour A., Merhof D. 2024. Medical image segmentation review: The success of u-net. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(12): 10076–10095. DOI: 10.1109/TPAMI.2024.3435571.
Cheng X., Chen J., Li J., Yin J., Cheng Q., Chen Z., Zhou G. 2025. Enhanced DeepLabV3+ with OBIA and Lightweight Attention for Accurate and Efficient Tree Species Classification in UAV Images. Sensors, 25(24): 7501. DOI: 10.3390/s25247501.
Chicco D., Jurman G. 2023. The Matthews correlation coefficient (MCC) should replace the ROC AUC as the standard metric for assessing binary classification. BioData mining, 16(1): 4. DOI: 10.1186/S13040-023-00322-4.
El Sakka M., Ivanovici M., Chaari L., Mothe J. 2025. A Review of CNN Applications in Smart Agriculture Using Multimodal Data. Sensors, 25(2): 472. DOI: 10.3390/s25020472.
Horuz C.C., Kasenbacher G., Higuchi S., Kairat S., Stoltz J., Pesl M., Otte S. 2025. The resurrection of the relu. arXiv, 2505: 22074. DOI: 10.48550/arXiv.2505.22074.
Jiangtao W., Ruhaiyem N.I.R., Panpan F. 2025. A comprehensive review of U‐Net and its variants: advances and applications in medical image segmentation. IET Image Processing, 19(1): e70019. DOI: 10.1049/ipr2.70019.
Kawaguchi K., Deng Z., Ji X., Huang J. 2023. How does information bottleneck help deep learning? International conference on machine learning: 16049–16096. DOI: 10.48550/arXiv.2305.18887.
Khan A., Rauf Z., Sohail A. et al. 2023. A survey of the vision transformers and their CNN-transformer based variants. Artif Intell Rev, 56 (3): 2917–2970. DOI: 10.1007/s10462-023-10595-0.
Kim J.W., Khan A.U., Banerjee I. 2025. Systematic Review of Hybrid Vision Transformer Architectures for Radiological Image Analysis. J Digit Imaging. Inform. med.: 38, 3248–3262. DOI: 10.1007/s10278-024-01322-4.
Li M., Guo Y., Guo W., Qiao H., Shi L., Liu Y., Wang Q. 2025. Wheat Powdery Mildew Severity Classification Based on an Improved ResNet34 Model. Agriculture, 15(15): 1580. DOI: 10.3390/agriculture15151580.
Li P., Zhou J., Xu X. 2025. Real-time image semantic segmentation based on improved deeplabv3+ network. Big Data and Cognitive Computing, 9(6): 152. DOI: 10.3390/bdcc9060152.
Li W., Zheng B., Shen Q., Shi X., Luo K., Yao Y., Wei Q. 2025. Adaptive window adjustment with boundary DoU loss for cascade segmentation of anatomy and lesions in prostate cancer using bpMRI. Neural Networks, 181: 106831. DOI: 10.1016/j.neunet.2024.106831.
Li Y., Zhao W., Dang B., Wang W. 2025. Research on Brain Tumor Classification Method Based on Improved ResNet34 Network. arXiv, 2512: 03751. DOI: 10.48550/arXiv.2512.03751.
Lufian C., Jonathan H., Lo Y.F., Iswanto I.A. 2025. Mask R-CNN with HRNet Backbone for Person Detection Evaluation on COCO Using mIoU and Threshold Optimization. 2025 8th International Conference on Information and Communications Technology (ICOIACT): 60–65. DOI: 10.1109/ICOIACT67584.2025.11344893.
Luo X., Cai Z., Shao B., Wang Y. 2024. Unified-IoU: For high-quality object detection. arXiv, 2408: 06636. DOI: 10.48550/arXiv.2408.06636.
Nemavhola A., Chibaya C., Viriri S. 2025. A Systematic Review of CNN Architectures, Databases, Performance Metrics, and Applications in Face Recognition. Information, 16(2): 107. DOI: 10.3390/info16020107.
Qiu Q., Zhu T., Gong H., Chen L., Ning H. 2025. Relu-kan: New kolmogorov-arnold networks that only need matrix addition, dot multiplication, and relu. 2025 IEEE Smart World Congress (SWC): 1686-1694. DOI: 10.1109/SWC65939.2025.00262.
Wang Y., Deng Y., Zheng Y., Chattopadhyay P., Wang L. 2025. Vision Transformers for Image Classification: A Comparative Survey. Technologies, 13(1): 32. DOI: 10.3390/technologies13010032.
Wen K., Chu J., Chen J., Chen Y., Cai J. 2022. MO SiamRPN with weight adaptive joint MIoU for UAV visual localization. Remote Sensing, 14(18): 4467. DOI: 10.3390/rs14184467.
Wu D., Guo Z., Li A., Yu C., Gao C., Sang N. 2023. Conditional boundary loss for semantic segmentation. IEEE Transactions on Image Processing, 32: 3717–3731. DOI: https://doi.org/10.1109/TIP.2023.3290519.
Yang J., Lu Y., Zhang Z., Wei J., Shang J., Wei C., Tang W., Chen J. 2025. A Deep Learning Method Coupling a Channel Attention Mechanism and Weighted Dice Loss Function for Water Extraction in the Yellow River Basin. Water, 17(4): 478. DOI: 10.3390/w17040478.
Yang T., Ying Y. 2022. AUC maximization in the era of big data and AI: A survey. ACM computing surveys, 55(8), 1 – 37. DOI: 10.1145/3554729.
Zheng Y., Tian B., Yu S., Yang X., Yu Q., Zhou J., Wang L. 2025. Adaptive boundary-enhanced Dice loss for image segmentation. Biomedical signal processing and control, 106: 107741. DOI: https://doi.org/10.1016/j.bspc.2025.107741.
Abstract views: 0
Share
Published
How to Cite
Issue
Section
Copyright (c) 2026 Economics. Information Technologies
This work is licensed under a Creative Commons Attribution 4.0 International License.
