Fine-Tuned AlexNet for Roof Shape Classification in Uzbekistan: a Transfer Learning Approach
DOI:
https://doi.org/10.71310/pcam.3_67.2025.12Keywords:
roof shape classification, convolutional neural networks, transfer learning, AlexNet, satellite imageryAbstract
This study investigates the classification of rooftop shapes using convolutional neural networks (CNNs), with a particular focus on regional adaptation through transfer learning. Initial training of the AlexNet architecture utilizes a publicly available Zenodo dataset comprising satellite imagery of flat, gabled, and hipped roofs. To address generalizability constraints in specific geographic contexts, a custom dataset of Uzbekistan rooftops, sourced from OpenStreetMap and high-resolution Mapbox Static Images API tiles, enables fine-tuning. Experimental outcomes reveal enhanced classification accuracy for flat, gabled, and hipped roofs, underscoring the efficacy of transfer learning in mitigating domain shift due to architectural and environmental variations. Integration of open-source geospatial tools with transfer learning offers a replicable framework for addressing geographic bias in rooftop shape classification, adaptable to other underrepresented regions.
References
Krizhevsky A., Sutskever I., Hinton G.E. 2012. Imagenet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 25,
Roof shape classification dataset https://zenodo.org/record/3986721,
Roof shapes of Uzbekistan dataset, https://drive.google.com/file/d/1jccNLjz3Tb3oMrrgb_HGHl_TIJprAq72/view?usp=sharing, OSM
OpenStreetMap contributors. Planet dump retrieved from https://planet.openstreetmap.org, n.d.
OpenStreetMap contributors 2007. Planet dump retrieved from https://planet.openstreetmap.org. https://www.openstreetmap.org
Chollet F. 2017. Deep Learning with Python. Manning Publications. Greenwich.
Simonyan K. and Zisserman A. 2015. Very deep convolutional networks for large-scale image recognition. volume arXiv:1409.1556, https://arxiv.org/abs/1409.1556
Yuldashev S. 2025. Roof Shape Classification in Uzbekistan Using Fine-tuned AlexNet. Google Colab Notebook, https://colab.research.google.com/drive/1tlQBzS7ALSqkhe12q8yYW_1AGHmUTSla?usp=sharing
He K., Zhang X., Ren S., Sun J. 2016. Deep residual learning for image recognition. Proc. IEEE Conf. Comput. Vis. Pattern Recognit. – P. 70–778.
Pritt M., Chern G. 2020. Satellite image classification with deep learning. arXiv preprint,
Simonyan K. and Zisserman A. 2015. Very deep convolutional networks for large-scale image recognition. arXiv preprint. arXiv:1409.1556,
He K., Zhang X., Ren S. and Sun J. 2016. Deep residual learning for image recognition. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., – P. 770–778.
Chroni A., Vasilakos C., Christaki M. and Soulakellis N. 2024. Fusing multispectral and lidar data for cnn-based semantic segmentation in semi-arid mediterranean environments. Remote Sensing, 16(15): 2729,
Chollet 2017. Deep Learning with Python. Manning Publications, Greenwich, CT,
Ni J., Chen J., Wu Y., Chen Z., Liang M. 2022. Method to determine the centroid of non-homogeneous polygons based on suspension theory. ISPRS Int. J. Geo-Inf., 11(4): 233,
Zhao Y., Li X., Wang Z., Chen H. and Liu J. 2025. Deep-learning-based evaluation of rooftop photovoltaic deployment in tianjin, china. ISPRS International Journal of Geo-Information, 14(3):101,
Jie Gu, Zhen Xie, Jiawei Zhang, and Xiaohong He. 2024. Advances in rapid damage identification methods for post-disaster regional buildings based on remote sensing images: A survey. Buildings, 14(4): 898,
Haoyu Hao, Sreeharsha Baireddy, Ethan R. Bartusiak, Robert T. Collins, and Michael K. Hinders. 2020. An attention-based system for damage assessment using satellite imagery. arXiv preprint arXiv:2004.06643
Jeremy D. Castagno and Ella M. Atkins. 2018. Roof shape classification from lidar and satellite image data. Sensors, 18(11):3960,
John E. Vargas-Muñoz et al. 2019. Correcting rural building annotations in openstreetmap using convolutional neural networks. arXiv preprint arXiv:1901.08190,
Shuochuan Meng, Mohammad Hesam Soleimani-Babakamali, and Ertugrul Taciroglu. 2023. Au- tomatic roof type classification through machine learning for regional wind risk assessment. arXiv preprint arXiv:2305.17315,
He K., Zhang X., Ren S., and J. Sun. 2016. Deep residual learning for image recognition. In Proc. IEEE Conf. Comput. Vis. Pattern Recognit., – P. 770–778.
Robins Kaplan LLP. 2023. Ai’s impact on property insurance cov- erage. https://www.robinskaplan.com/newsroom/insights/ ais-impact-on-property-insurancecoverage, Accessed: 2025-05-24.
Yuldashev 2025. Roof shape classification in Uzbekistan using fine-tuned alexnet Google Colab Notebook.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 S.U. Yuldashev
This work is licensed under a Creative Commons Attribution 4.0 International License.
