Az Örnekli Öğrenme Yöntemiyle Ağaçlık Alan Bölütlemesi

Öz

Ağaçlık alanların bölütlenmesi, çevresel sürdürülebilirlik takibi ve haberleşme altyapı planlaması gibi alanlarda kritik bir role sahiptir; ancak tam denetimli derin öğrenme modellerinin piksel tabanlı yoğun veri etiketleme gereksinimi, maliyet, zaman ve uygulanabilirlik açısından önemli kısıtlar oluşturmaktadır. Bu çalışma söz konusu kısıtlamaları aşmak amacıyla yüksek çözünürlüklü hava görüntüleri üzerinde ağaçlık alan tespiti için prototip ağları tabanlı bir az örnekli öğrenme yaklaşımı sunmaktadır. LandCover.ai veri kümesi üzerinde gerçekleştirilen sistematik deneylerde, evrişimli ve dönüştürücü tabanlı sekiz farklı öznitelik omurgası dondurulmuş ağırlıklarla karşılaştırılmış ve ardından en başarılı evrişimli mimariler üzerinde katman bazlı ince ayar stratejileri değerlendirilmiştir. Deneysel sonuçlar, önerilen yöntemin en yüksek başarımı dondurulmuş koşuldaki dönüştürücü tabanlı mimari ile elde ettiğini, evrişimli mimarilerde ise kontrollü katman açmanın başarımı daha da artırabildiğini göstermektedir. Önerilen yöntem, yoğun piksel düzeyinde etiketleme gerektirmeden yalnızca beş etiketli destek örneğiyle kararlı ve tutarlı bir bölütleme başarımı sergilemiştir.

Anahtar Kelimeler

• Az Örnekli Öğrenme
• Anlamsal Bölütleme
• Prototip Ağları
• Hava Görüntüleri
• Veri Verimliliği

Woodland Segmentation Using Few-shot Learning

Öz

Woodland segmentation plays a critical role in environmental sustainability monitoring and telecommunications infrastructure planning; however, fully supervised deep learning models impose significant constraints due to their intensive pixel-level annotation requirements. This study proposes a Prototypical Networks-based few-shot learning approach for woodland detection in high-resolution aerial imagery. Systematic experiments on the LandCover.ai dataset compared eight convolutional and transformer-based backbone architectures using frozen weights, followed by an evaluation of layer-wise fine-tuning strategies on the best-performing convolutional backbones. Experimental results demonstrate that the transformer-based architecture achieves the highest performance under frozen conditions, while controlled layer unfreezing on convolutional architectures further improves segmentation accuracy. The proposed method achieves stable and consistent segmentation performance using only five labeled support examples without requiring dense pixel-level annotation.

Anahtar Kelimeler

• Few-Shot Learning
• Semantic Segmentation
• Prototypical Networks
• Aerial Imagery
• Data Efficiency

Kaynakça

1. [1] Boguszewski, A., Batorski, D., Ziemba-Jankowska, N., Dziedzic, T., & Zambrzycka, A. (2021). LandCover. ai: ataset for automatic mapping of buildings, woodlands, water and roads from aerial imagery. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 1102–1110).
2. [2] Petrov, M., Pandilova, E., Dimitrovski, I., Trajanov, D., Spasev, V., & Kitanovski, I. (2026). Few-shot semantic segmentation in remote sensing: A review on definitions, methods, datasets, advances and future trends. Remote Sensing, 18(4), 637.
3. [3] Sun, X.,Wang, B.,Wang, Z., Li, H., Li, H., & Fu, K. (2021). Research progress on few-shot learning for remote sensing image interpretation. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14, 2387–2402.
4. [4] Darici, M. B., Gumuslu, E., & Serifoglu, M. O. (2025, June). Woodland Segmentation: A Comprehensive Analysis on China Region. In 2025 33rd Signal Processing and Communications Applications Conference (SIU) (pp. 1–4). IEEE.
5. [5] Chen, L. C., Papandreou, G., Kokkinos, I., Murphy, K., & Yuille, A. L. (2017). Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. IEEE Transactions on Pattern Analysis and Machine Intelligence, 40(4), 834–848.
6. [6] Chen, J., Wang, X., Hong, L., & Liu, M. (2024). Cross-domain few-shot segmentation for remote sensing image based on task augmentation and feature disentanglement. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 17, 9360–9375.
7. [7] Dong, N., & Xing, E. P. (2018, September). Fewshot semantic segmentation with prototype learning. In BMVC (Vol. 3, No. 4, p. 4).
8. [8] Snell, J., Swersky, K., & Zemel, R. (2017). Prototypical networks for few-shot learning. Advances in Neural Information Processing Systems, 30.

9. [9] Tang, H., Jia, Y., Cheng, J., Mu, Y., Wu, Y., & Yao, X. (2026). GlocalDualNet: Disentangling Scale and Representation for Few-Shot Remote Sensing Segmentation. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.
10. [10] Long, J., Shelhamer, E., & Darrell, T. (2015). Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 3431– 3440).
11. [11] Ronneberger, O., Fischer, P., & Brox, T. (2015, October). U-Net: Convolutional networks for biomedical image segmentation. In International Conference on Medical Image Computing and Computer- Assisted Intervention (pp. 234–241). Springer.
12. [12] He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 770–778).
13. [13] Zhao, H., Shi, J., Qi, X., Wang, X., & Jia, J. (2017). Pyramid scene parsing network. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 2881–2890).
14. [14] Lin, T. Y., Doll´ar, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2017). Feature pyramid networks for object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 2117–2125). [15] Chen, L. C., Zhu, Y., Papandreou, G., Schroff, F., & Adam, H. (2018). Encoder-decoder with atrous separable convolution for semantic image segmentation. In Proceedings of the European Conference on Computer Vision (ECCV) (pp. 801–818).
15. [16] Tian, Z., Zhao, H., Shu, M., Yang, Z., Li, R., & Jia, J. (2020). Prior guided feature enrichment network for few-shot segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(2), 1050–1065.
16. [17] Vinyals, O., Blundell, C., Lillicrap, T., & Wierstra, D. (2016). Matching networks for one shot learning. Advances in Neural Information Processing Systems, 29. [18] Finn, C., Abbeel, P., & Levine, S. (2017, July). Model-agnostic meta-learning for fast adaptation of deep networks. In International Conference on Machine Learning (pp. 1126–1135). PMLR.
17. [19] Shaban, A., Bansal, S., Liu, Z., Essa, I., & Boots, B. (2017). One-shot learning for semantic segmentation. arXiv preprint arXiv:1709.03410.
18. [20] Wang, K., Liew, J. H., Zou, Y., Zhou, D., & Feng, J. (2019). PANet: Few-shot image semantic segmentation with prototype alignment. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 9197–9206).
19. [21] Zhang, C., Lin, G., Liu, F., Yao, R., & Shen, C. (2019). CANet: Class-agnostic segmentation networks with terative refinement and attentive fewshot learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 5217–5226).
20. [22] Min, J., Kang, D., & Cho, M. (2021). Hypercorrelation squeeze for few-shot segmentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 6941–6952).
21. [23] Fan, Q., Pei, W., Tai, Y. W., & Tang, C. K. (2022, October). Self-support few-shot semantic segmentation. In European Conference on Computer Vision (pp. 701–719). Springer.
22. [24] Broni-Bediako, C., Xia, J., Song, J., Chen, H., Siam, M., & Yokoya, N. (2024). Generalized few-shot semantic segmentation in remote sensing: Challenge and benchmark. IEEE Geoscience and Remote Sensing Letters, 21, 1-5.
23. [25] Berka, A., Hajji, M. E., Canals, R., Es-saady, Y., & Hafiane, A. (2025). Enhancing DeepLabV3+ to fuse aerial and satellite images for semantic segmentation. arXiv preprint arXiv:2503.22909.
24. [26] Chollet, F. (2017). Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 1251–1258).
25. [27] Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., ... & Guo, B. (2021). Swin Transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 10012–10022).
26. [28] Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., ... & Houlsby, N. (2020). An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929.
27. [29] Deng, J., Dong, W., Socher, R., Li, L. J., Li, K., & Fei-Fei, L. (2009, June). Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition (pp. 248-255). Ieee.
28. [30] Lin, T. Y., Goyal, P., Girshick, R., He, K., & Doll´ar, P. (2017). Focal loss for dense object detection. In Proceedings of the IEEE International Conference on Computer Vision (pp. 2980–2988).
29. [31] Loshchilov, I., & Hutter, F. (2017). Decoupled weight decay regularization. arXiv preprint ar- Xiv:1711.05101.
30. [32] Loshchilov, I., & Hutter, F. (2016). SGDR: Stochastic gradient descent with warm restarts. arXiv preprint arXiv:1608.03983.
31. [33] Powers, D. M. (2020). Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlation. arXiv preprint ar- Xiv:2010.16061.