İnce Taneli Hava Aracı Sınıflandırmasında Hiyerarşi-Bilinçli Çapraz Seviye Dikkat Mekanizması

Öz

Görsel tanıma sistemlerinde ince taneli sınıflandırma problemleri, birbirine benzer alt kategorilerin ayırt edilmesini gerektirmekte ve bu durum özellikle hava aracı tanıma gibi alanlarda kritik zorluklar ortaya çıkarmaktadır. Mevcut yaklaşımlar genellikle hiyerarşik yapıyı göz ardı ederek düz sınıflandırma stratejileri kullanmaktadır. Bu araştırmada, çok seviyeli taksonomi bilgisini etkin biçimde kullanan özgün bir dikkat mekanizması önerilmektedir. Geliştirilen Hiyerarşi-Bilinçli Çapraz Seviye Dikkat (HACLA) mimarisi, üst kategorilerden alt kategorilere bilgi aktarımını dinamik olarak gerçekleştirmektedir. FGVC-Aircraft referans veri kümesi üzerinde yapılan kapsamlı deneyler, önerilen yöntemin varyant düzeyinde %89,31 ± 0,97 doğruluk elde ettiğini ve düz sınıflandırma yaklaşımlarına kıyasla %13,67 iyileşme sağladığını göstermektedir. Elde edilen bulgular, hiyerarşik bilgi aktarımının ince taneli görsel tanıma problemlerinde belirleyici bir rol oynadığını göstermektedir. 

Anahtar Kelimeler

• Derin Öğrenme
• Dikkat Mekanizması
• Hava Aracı Tanıma
• Hiyerarşik Öğrenme
• İnce Taneli Görsel Sınıflandırma

Hierarchy-Aware Cross-Level Attention Mechanism for Fine-Grained Aircraft Classification

Öz

Fine-grained classification problems in visual recognition systems require distinguishing highly similar subcategories, posing critical challenges particularly in domains such as aircraft recognition. Existing approaches typically employ flat classification strategies, disregarding the inherent hierarchical structure. In this study, a novel attention mechanism that effectively leverages multi-level taxonomy information is proposed. The developed Hierarchy-Aware Cross-Level Attention (HACLA) architecture dynamically performs information transfer from parent categories to child categories. Extensive experiments conducted on the FGVC-Aircraft benchmark dataset demonstrate that the proposed method achieves 89,31 ± 0,97% accuracy at the variant level and provides a 13,67% improvement compared to flat classification approaches. The findings indicate that hierarchical information transfer plays a decisive role in fine-grained visual recognition problems. 

Anahtar Kelimeler

• Deep Learning
• Attention Mechanism
• Aircraft Recognition
• Hierarchical Learning
• Fine-Grained Visual Classification

Kaynakça

1. Chang, D., Pang, K., Zheng, Y., Ma, Z., Song, Y. Z., & Guo, J. (2021). Your "flamingo" is my "bird": Fine-grained, or not. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 11476–11485. https://doi.org/10.48550/arXiv.2011.09040
2. Cui, Y., Song, Y., Sun, C., Howard, A., & Belongie, S. (2018). Large scale fine-grained categorization and domain-specific transfer learning. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). https://doi.org/10.48550/arXiv.1806.06193
3. Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009). ImageNet: A large-scale hierarchical image database. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 248–255. https://doi.org.10.1109/CVPR.2009.5206848
4. Ding, Y., Ma, Z., Wen, S., Xie, J., Chang, D., Si, Z., ... & Ling, H. (2021). AP-CNN: Weakly supervised attention pyramid convolutional neural network for fine-grained visual classification. IEEE Transactions on Image Processing, 30, 2826–2836. https://doi.org. 10.1109/TIP.2021.3055617
5. Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., ... & Houlsby, N. (2021). An image is worth 16x16 words: Transformers for image recognition at scale (arXiv Preprint arXiv:2010.11929). arXiv. https://arxiv.org/abs/2010.11929
6. Du, R., Chang, D., Bhunia, A., Xie, J., Ma, Z., Song, Y., & Guo, J. (2020). Fine-grained visual classification via progressive multi-granularity training of jigsaw patches. European Conference on Computer Vision (ECCV). https://doi.org/10.48550/arXiv.2003.03836
7. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 770–778. https://doi.org/10.1109/CVPR.2016.90
8. He, J., Chen, J.-N., Liu, S., Kortylewski, A., Yang, C., Bai, Y., & Wang, C. (2022). TransFG: A transformer architecture for fine-grained recognition. AAAI Conference on Artificial Intelligence, 36(1), 852–860. https://doi.org/10.48550/arXiv.2103.07976

9. Huang, W., Chen, E., Liu, Q., Chen, Y., Huang, Z., Liu, Y., ... & Wang, S. (2019). Hierarchical multi-label text classification: An attention-based recurrent network approach. In Proceedings of the 28th ACM International Conference on Information and Knowledge Management (pp. 1051–1060). https://doi.org/10.1145/3357384.3357885
10. Ke, X., Cai, Y., Chen, B., Liu, H., & Guo, W. (2025). Multi-granularity interaction and feature recombination network for fine-grained visual classification. Pattern Recognition, 166(9), Article 111632. https://doi.org/10.1016/j.patcog.2025.111632
11. Kullback, S., & Leibler, R. A. (1951). On information and sufficiency. The Annals of Mathematical Statistics, 22(1), 79–86. https://doi.org/10.1214/aoms/1177729694
12. Lin, T.-Y., RoyChowdhury, A., & Maji, S. (2015). Bilinear CNN models for fine-grained visual recognition. IEEE International Conference on Computer Vision (ICCV), 1449–1457. https://doi.org/10.1109/ICCV.2015.170
13. Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., & Guo, B. (2021). Swin Transformer: Hierarchical vision transformer using shifted windows. IEEE/CVF International Conference on Computer Vision (ICCV), 10012–10022. https://doi.org/10.48550/arXiv.2103.14030
14. Liu, Z., Mao, H., Wu, C., Feichtenhofer, C., Darrell, T., & Xie, S. (2022). A ConvNet for the 2020s. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). https://doi.org/10.48550/arXiv.2201.03545
15. Liu, Y., Yang, L., & Wang, Y. (2024). Hierarchical fine-grained visual classification leveraging consistent hierarchical knowledge. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (pp. 279–295). Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-70341-6_17
16. Loshchilov, I., & Hutter, F. (2019). Decoupled weight decay regularization. International Conference on Learning Representations (ICLR). https://doi.org/10.48550/arXiv.1711.05101
17. Maji, S., Rahtu, E., Kannala, J., Blaschko, M., & Vedaldi, A. (2013). Fine-grained visual classification of aircraft (arXiv Preprint arXiv:1306.5151). https://arxiv.org/abs/1306.5151
18. Miao, Z., Zhao, X., Wang, J., Li, Y., & Li, H. (2021). Complemental attention multi-feature fusion network for fine-grained classification. IEEE Signal Processing Letters, 28, 1983–1987. https://doi.org/10.1109/LSP.2021.3114622
19. Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., ... & Chintala, S. (2019). Pytorch: An imperative style, high-performance deep learning library. Advances in Neural Information Processing Systems, 32. https://doi.org/10.48550/arXiv.1912.01703
20. Silla, C., & Freitas, A. (2011). A survey of hierarchical classification across different application domains. Data Mining and Knowledge Discovery, 22, 31–72. https://doi.org/10.1007/s10618-010-0175-9
21. Van der Maaten, L., & Hinton, G. (2008). Visualizing data using t-SNE. Journal of Machine Learning Research, 9(86), 2579–2605.
22. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems (NeurIPS), 30. https://doi.org/10.48550/arXiv.1706.03762
23. Wang, R., Zou, C., Zhang, W., Zhu, Z., & Jing, L. (2023). Consistency-aware feature learning for hierarchical fine-grained visual classification. ACM International Conference on Multimedia (MM '23), 2326–2334. https://doi.org/10.1145/3581783.3612234
24. Wehrmann, J., Cerri, R., & Barros, R. (2018). Hierarchical multi-label classification networks. International Conference on Machine Learning (ICML), 5075–5084.
25. Wei, X. S., Wu, J., & Cui, Q. (2019). Deep learning for fine-grained image analysis: A survey. arXiv preprint arXiv:1907.03069. https://doi.org/10.48550/arXiv.1907.03069
26. Wightman, R. (2019). PyTorch Image Models. GitHub. https://github.com/rwightman/pytorch-image-models
27. Woo, S., Debnath, S., Hu, R., Chen, X., Liu, Z., Kweon, I. S., & Xie, S. (2023). ConvNeXt V2: Co-designing and scaling ConvNets with masked autoencoders. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). https://doi.org/10.48550/arXiv.2301.00808
28. Yun, S., Han, D., Oh, S. J., Chun, S., Choe, J., & Yoo, Y. (2019). CutMix: Regularization strategy to train strong classifiers with localizable features. IEEE/CVF International Conference on Computer Vision (ICCV), 6023–6032. https://doi.org/10.48550/arXiv.1905.04899
29. Zhang, N., Donahue, J., Girshick, R., & Darrell, T. (2014). Part-based R-CNNs for fine-grained category detection. European Conference on Computer Vision (ECCV), 834–849. https://doi.org/10.48550/arXiv.1407.3867
30. Zhang, X., Zhang, K., Zhao, W., Luo, H., Zhong, S., Tang, L., ... & Fan, J. (2023). A benchmark dataset and approach for fine-grained visual categorization in complex scenes. Digital Signal Processing, 137, 104033. https://doi.org/10.1016/j.dsp.2023.104033
31. Zheng, H., Fu, J., Mei, T., & Luo, J. (2017). Learning multi-attention convolutional neural network for fine-grained image recognition. IEEE International Conference on Computer Vision (ICCV), 5209–5217. https://doi.org/ 10.1109/ICCV.2017.557