BEAU-Net: Boundary And Edge Aware U-Net For Improved Building Segmentation

Akil Ahmad Taki; Md. Rifat Ur Rahman; Mohammed Imamul Hassan Bhuiyan; Md Sultan Mahmud

doi:10.29284/2vcbx071

Authors

Akil Ahmad Taki Department of EEE, University of Asia Pacific, Dhaka, Bangladesh
Md. Rifat Ur Rahman Department of EEE, Bangladesh University of Engineering and Technology,Dhaka Bangladesh
Mohammed Imamul Hassan Bhuiyan Department of EEE, Bangladesh University of Engineering and Technology,Dhaka Bangladesh
Md Sultan Mahmud Department of EEE, University of Asia Pacific, Dhaka, Bangladesh

DOI:

https://doi.org/10.29284/2vcbx071

Keywords:

Building footprint extraction, U-Net architecture, Remote sensing imagery, High-resolution satellite data, Boundary delineation, Edge recognition, Boundary attention module (BAM), Edge attention module (EAM).

Abstract

Extracting building footprints automatically from high-resolution remote sensing imagery has emerged as a focal area of active investigation. Over the past few years, deep learning techniques, especially convolutional neural networks (CNNs), have delivered encouraging outcomes in this domain. Nevertheless, existing leading-edge architectures continue to face difficulties when dealing with building boundaries, abrupt edge transitions, and intricate geometric configurations. To tackle these issues, we introduce Boundary and Edge Aware U-Net (BEAU-Net), an attention-driven segmentation framework built upon the UNet backbone. The BAM component concentrates on features containing abundant boundary cues. This strengthens the network’s ability to separate buildings from neighboring structures, which proves especially valuable in dense urban settings where buildings are closely spaced. The EAM component accentuates fine-grained features including precise edges and elaborate geometric patterns. This allows the network to represent intricate building geometries and identify sudden shifts in edge characteristics. Comprehensive experiments were carried out on the Massachusetts Building, WHU Building, and INRIA Aerial Image Labeling datasets to confirm the efficacy of our proposed framework. The findings reveal notable gains in both IoU and F1 metrics, particularly in scenarios involving tight boundaries and complex structural forms.

References

[1] C. Iovan, D. Boldo, and M. Cord, “Detection, characterization, and modeling vegetation in urban areas from high-resolution aerial imagery,”IEEE Journal of Selected Topics in Applied Earth Observations and

Remote Sensing, vol. 1, no. 3, pp. 206–213, 2008.

[2] V. Badrinarayanan, A. Kendall, and R. Cipolla, “Segnet: A deep convolutional encoder-decoder architecture for image segmentation,” IEEE transactions on pattern analysis and machine intelligence, vol. 39,

no. 12, pp. 2481–2495, 2017.

[3] O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” in Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings,

part III 18. Springer, 2015, pp. 234–241.

[4] Z. Zhou, M. M. Rahman Siddiquee, N. Tajbakhsh, and J. Liang,“UNet++: A Nested U-Net Architecture for Medical Image Segmentation,” in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: 4th International Workshop,DLMIA 2018, and 8th International Workshop, ML-CDS 2018, Held in Conjunction with MICCAI 2018, Granada, Spain, September 20, 2018,Proceedings 4. Springer, 2018, pp. 3–11.

[5] H. Huang, L. Lin, R. Tong, H. Hu, Q. Zhang, Y. Iwamoto, X. Han, Y.-W. Chen, and J. Wu, “UNet 3+: A Full-Scale Connected UNet for Medical Image Segmentation,” in ICASSP 2020-2020 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2020, pp. 1055–1059.

[6] Y. Li, Y. Li, X. Zhu, H. Fang, and L. Ye, “A method for extracting buildings from remote sensing images based on 3DJA-UNet3+,” Scientific Reports, vol. 14, no. 1, p. 19067, 2024.

[7] F. I. Diakogiannis, F. Waldner, P. Caccetta, and C. Wu, “ResUNet-a: A deep learning framework for semantic segmentation of remotely sensed data,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 162,

pp. 94–114, 2020.

[8] S. El Hajjar, H. Kassem, F. Abdallah, and H. Omrani, “Enhancing building segmentation by deep multiview classification for advancing sustainable urban development,” Journal of Building Engineering, vol. 83, p. 108421, 2024.

[9] M. Yu, X. Chen, W. Zhang, and Y. Liu, “AGs-Unet: Building Extraction Model for High Resolution Remote Sensing Images Based on Attention Gates U Network,” Sensors, vol. 22, no. 8, p. 2932, 2022.

[10] H. Zhao, H. Zhang, and X. Zheng, “A multiscale attention-guidedUNet++ with edge constraint for building extraction from high spatial resolution imagery,” Applied Sciences, vol. 12, no. 12, p. 5960, 2022.

[11] J. Lei, X. Liu, H. Yang, Z. Zeng, and J. Feng, “Dual Hybrid Attention Mechanism-Based U-Net for Building Segmentation in Remote Sensing Images,” Applied Sciences, vol. 14, no. 3, p. 1293, 2024.

[12] R. Wang, S. Chen, C. Ji, J. Fan, and Y. Li, “Boundary-aware context neural network for medical image segmentation,” Medical image analysis, vol. 78, p. 102395, 2022.

[13] X. Sun, A. Shi, H. Huang, and H. Mayer, “BAS4 Net: BoundaryAware Semi-Supervised Semantic Segmentation Network for Very High Resolution Remote Sensing Images ,” IEEE Journal of Selected Topics

in Applied Earth Observations and Remote Sensing, vol. 13, pp. 5398–

5413, 2020.

[14] Z. Nong, X. Su, Y. Liu, Z. Zhan, and Q. Yuan, “Boundary-aware dualstream network for vhr remote sensing images semantic segmentation,”IEEE Journal of Selected Topics in Applied Earth Observations and

Remote Sensing, vol. 14, pp. 5260–5268, 2021.

[15] A. Li, L. Jiao, H. Zhu, L. Li, and F. Liu, “Multitask semantic boundary awareness network for remote sensing image segmentation,” IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1–14, 2021.

[16] F. Milletari, N. Navab, and S.-A. Ahmadi, “V-net: Fully convolutional neural networks for volumetric medical image segmentation,” in 2016 fourth international conference on 3D vision (3DV). Ieee, 2016, pp.

565–571.

[17] M. E. Omati and F. T. Mahmoudi, “Efficiency analysis of itn loss function for deep semantic building segmentation,” Earth Science Informatics, pp. 1–15, 2024.

[18] L. Liu, J. Zhang, R. He, Y. Liu, Y. Wang, Y. Tai, D. Luo, C. Wang, J. Li, and F. Huang, “Learning by analogy: Reliable supervision from transformations for unsupervised optical flow estimation,” in 2020

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 6488–6497.

[19] L.-C. Chen, “Rethinking atrous convolution for semantic image segmentation,” arXiv preprint arXiv:1706.05587, 2017.

[20] S. Xie and Z. Tu, “Holistically-nested edge detection,” in Proceedings of the IEEE international conference on computer vision, 2015, pp. 1395–1403.

[21] Y. Liu, M.-M. Cheng, X. Hu, K. Wang, and X. Bai, “Richer convolutional features for edge detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 3000–

3009.

[22] S. Chen, W. Shi, M. Zhou, M. Zhang, and Z. Xuan, “CGSANet: A contour-guided and local structure-aware encoder–decoder network for accurate building extraction from very high-resolution remote sensing

imagery,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 15, pp. 1526–1542, 2021.

[23] H. Zhao, J. Shi, X. Qi, X. Wang, and J. Jia, “Pyramid scene parsing network,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 2881–2890.

[24] L.-C. Chen, Y. Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoderdecoder with atrous separable convolution for semantic image segmentation,” in Proceedings of the European conference on computer vision

(ECCV), 2018, pp. 801–818.

[25] Y. Wu, P. Zhao, F. Wang, M. Zhou, S. Geng, and D. Zhang, “A prior-guided dual branch multi-feature fusion network for building segmentation in remote sensing images,” Buildings, vol. 14, no. 7, p.2006, 2024.

[26] V. Mnih, “Machine learning for aerial image labeling,” in PhD Thesis, University of Toronto, 2013.

[27] S. Ji, S. Wei, and M. Lu, “Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set,” IEEE Transactions on Geoscience and Remote Sensing, vol. 57, no. 1,

pp. 574–586, 2018.

[28] E. Maggiori, Y. Tarabalka, G. Charpiat, and P. Alliez, “A large-scale benchmark for semantic segmentation of very high resolution satellite images,” in Proceedings of the IEEE International Conference on Computer Vision Workshops, 2017, pp. 161–169.