Deep Learning Models Integrating Attention Mechanisms For Military Camouflaged Object Detection

Öz

: The Adaptive Camouflage Data Set (ACD1K) is a carefully curated collection of high-quality images developed to facilitate research in camouflage detection and segmentation tasks. This dataset is categorized into training, validation, and testing subsets, enabling comprehensive evaluation of deep learning models. Models including Attention U-Net, built upon the ResNet-50 architecture, and U-Net++, enhanced with attention mechanisms, were employed for robust feature extraction. Performance evaluation was carried out using common metrics such as accuracy, precision, recall, F1-score, and intersection over union. The Attention U-Net model, in conjunction with CLAHE preprocessing, Adamax optimizer, a learning rate of 1e-5, and a dropout rate of 0.2, achieved an accuracy of 96.88% and an intersection over union of 92.01%. Under similar experimental conditions, the Attention U-Net++ model using the Adam optimizer achieved an accuracy of 98.32% and an intersection over union of 82.09%. These findings highlight the effectiveness of CNN-based architectures in accurately identifying camouflaged objects within visually complex environments.

Anahtar Kelimeler

• Attention Mechanisms
• Deep Learning
• Image Segmentation
• Object Detection
• Adaptive Camouflage Dataset

Kamufle Askerî Nesnelerin Tespiti İçin Dikkat Mekanizmalarıyla Entegre Edilmiş Derin Öğrenme Modelleri

Öz

Adaptive Camouflage Veri Seti (ACD1K), kamuflaj tespiti ve segmentasyon görevlerine yönelik araştırmaları desteklemek amacıyla özenle derlenmiş, yüksek kaliteli görsellerden oluşan bir koleksiyondur. Bu veri seti, derin öğrenme modellerinin kapsamlı bir şekilde değerlendirilmesine olanak tanımak üzere eğitim, doğrulama ve test alt kümelerine ayrılmıştır. Özellikle ResNet-50 mimarisi üzerine inşa edilmiş Attention U-Net modeli ile dikkat mekanizmalarıyla geliştirilen U-Net++ modeli, güçlü özellik çıkarımı amacıyla kullanılmıştır. Modellerin performansı; doğruluk (accuracy), kesinlik (precision), duyarlılık (recall), F1-skoru ve IoU (intersection over union) gibi yaygın metrikler kullanılarak değerlendirilmiştir. CLAHE önişleme yöntemi, Adamax optimizasyon algoritması, 1e-5 öğrenme oranı ve %20 dropout oranı ile birlikte kullanılan Attention U-Net modeli, %96.88 doğruluk ve %92.01 IoU değeri elde etmiştir. Benzer deneysel koşullar altında, Adam optimizasyon algoritmasıyla eğitilen Attention U-Net++ modeli ise %98.32 doğruluk ve %82.09 IoU performansı göstermiştir. Bu bulgular, görsel olarak karmaşık ortamlarda kamufle edilmiş nesnelerin doğru şekilde tespiti açısından, CNN tabanlı mimarilerin etkinliğini açıkça ortaya koymaktadır.

Anahtar Kelimeler

• Dikkat Mekanizmaları
• Derin Öğrenme
• Görüntü Bölütleme
• Nesne Tespiti
• Uyarlanabilir Kamuflaj Veri Kümesi

Kaynakça

1. [1] K. Karthiga and A. Asuntha, “CAMOUFLAGE-Net: comprehensive advanced model for optimal camouflaged target detection and analysis using groundbreaking elements,” Signal, Image and Video Processing, 2025. doi: 10.1007/s11760-025-02928-5.
2. [2] B. Janakiramaiah et al., “Military object detection in defence using multi-level capsule networks,” Research Square Preprint RS-3210306, 2023. [Online]. Available: https://www.researchsquare.com/article/rs-3210306/v1.
3. [3] C. Guo and H. Huang, “Enhancing camouflaged object detection through contrastive learning and data augmentation techniques,” Engineering Applications of Artificial Intelligence, vol. 141, p. 109703, 2025. doi: 10.1016/j.engappai.2025.109703.
4. [4] X. Jiang et al., “MAGNet: a camouflage object detection network simulating the observation effect of magnifier,” Research Square Preprint RS- 976369, 2021. [Online]. Available: https://www.researchsquare.com/article/rs-976369/v1.
5. [5] B. Li, R. Zhou, L. Yang, Q. Wang and H. Chen, “MilDetr: detection transformer for military camouflaged target detection,” IEEE Access, vol. 12, art. no. 10450905, 2024. doi: 10.1109/ACCESS.2024.3366974.
6. [6] O. Ronneberger, P. Fischer and T. Brox, “U-Net: convolutional networks for biomedical image segmentation,” arXiv preprint arXiv:1505.04597, 2015.
7. [7] Z. Ahmed, S. A. Tanim, F. S. Prity, H. Rahman, and T. B. M. Maisha, “Improving biomedical image segmentation: an extensive analysis of U-Net for enhanced performance,” in Proc. ICETITE, Vellore, India, Feb. 2024.
8. [8] K. Türkarslan ve F. Hardalac, “Derin Öğrenme Yöntemleri Kullanılarak Havadan Elde Edilen Görüntüler Üzerinde Nesne Tespiti”, ECJSE, c. 9, sy. 4, ss. 1398–1410, 2022, doi: 10.31202/ecjse.1135509.

9. [9] T. Yuan et al., “CoFiNet: unveiling camouflaged objects with multi-scale feature integration,” SciSpace/Typeset.io, [Online]. Available: https://typeset.io/papers/cofinet-unveiling-camouflaged-objects-with-multi-scale-feature-oov8w5r71u. [Accessed: May 14, 2025].
10. [10] D. P. Naik and S. C. Tripathi, “Improving camouflage object detection using U-Net and VGG16,” SciSpace/Typeset.io, [Online]. Available: https://typeset.io/papers/improving-camouflage-object-detection-using-u-net-and-vgg16-x4vj1d3zrv. [Accessed: May 14, 2025].
11. [11] H. Anandan, V. Karthikeyan and R. Jaisankar, “U-Net based camouflaged object detection,” SciSpace/Typeset.io, [Online]. Available: https://typeset.io/papers/u-net-based-camouflaged-object-detection-28p6o029t7. [Accessed: May 14, 2025].
12. [12] Y. Zhang, J. Zhang, Y. Cao and H. Liu, “DMNet: diffusion model network for camouflaged object detection,” Electronics, vol. 13, no. 18, p. 3690, 2024. doi: 10.3390/electronics13183690.
13. [13] "Adaptive camouflaged dataset (ACD1K)", Kaggle, 2024. [Online]. Available: https://www.kaggle.com/datasets/5f06c6964edcaa5f421efa39144d101532e19410e1323d2a75757d9f34baaf87. [Accessed: 04-May-2025].
14. [14] R. Kamble and P. Rajarajeswari, “Revealing hidden patterns: a deep learning approach to camouflage detection,” International Journal of Computational Methods and Experimental Measurements, vol. 12, no. 1, pp. 97–105, Mar. 2024. doi: 10.18280/ijcmem.120111.
15. [15] R. J. Hyndman and A. B. Koehler, “Another look at measures of forecast accuracy,” International Journal of Forecasting, vol. 22, no. 4, pp. 679–688, Oct. 2006. doi: 10.1016/j.ijforecast.2006.03.001.
16. [16] X. Ji and S. Jiang, “Research on natural image semantic segmentation method based on improved U-Net model,” in Third International Conference on Electronics, Electrical and Information Engineering (ICEEIE 2023), SPIE, p. 87, Oct. 2023. doi: 10.1117/12.3009019.
17. [17] A. Buslaev, A. Parinov, E. Khvedchenya, V. I. Iglovikov and A. A. Kalinin, “Albumentations: fast and flexible image augmentations,” Information, vol. 11, no. 2, p. 125, 2020. doi: 10.3390/info11020125.
18. [18] A. Mishra, “Contrast limited adaptive histogram equalization (CLAHE) approach for enhancement of the microstructures of friction stir welded joints,” arXiv preprint arXiv:2109.00886, 2021.
19. [19] S. Henna and A. Reji, “A data augmented approach to transfer learning for Covid-19 detection,” arXiv preprint arXiv:2108.02870, 2021.
20. [20] A. Almansoori et al., “Image preprocessing with contrast-limited adaptive histogram equalization improves the segmentation performance of deep learning for the articular disk of the temporomandibular joint on MRI,” Computers in Biology and Medicine, vol. 152, p. 106334, 2023. doi: 10.1016/j.compbiomed.2022.106334.
21. [21] A. Zafar et al., “Bayesian optimization for CLAHE parameter tuning in medical image analysis,” Scientific Reports, vol. 14, no. 1, p. 5581, 2024. doi: 10.1038/s41598-024-55460-w.
22. [22] S. Sahu and A. Singh, “A comparative study on CLAHE and deep learning based enhancement techniques for low-contrast medical images,” Biomedical Signal Processing and Control, vol. 75, p. 103591, 2022. doi: 10.1016/j.bspc.2022.103591.
23. [23] K. He, X. Zhang, S. Ren and J. Sun, “Deep residual learning for image recognition,” in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, Jun. 2016, pp. 770–778. doi: 10.1109/CVPR.2016.90.
24. [24] S. Zagoruyko and N. Komodakis, “Wide residual networks,” arXiv preprint arXiv:1605.07146, 2016.
25. [25] J. Hu, L. Shen and G. Sun, “Squeeze-and-excitation networks,” in 2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA, Jun. 2018, pp. 7132–7141. doi: 10.1109/CVPR.2018.00744.
26. [26] S. Xie, R. Girshick, P. Dollár, Z. Tu and K. He, “Aggregated residual transformations for deep neural networks,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, Jul. 2017, pp. 1492–1500. doi: 10.1109/CVPR.2017.165.
27. [27] G. Huang, Z. Liu, L. van der Maaten and K. Q. Weinberger, “Densely connected convolutional networks,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, Jul. 2017, pp. 4700–4708. doi: 10.1109/CVPR.2017.243.
28. [28] Y. Sun, X. Wang and X. Tang, “Deeply learned face representations are sparse, selective, and robust,” in 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA, Jun. 2015, pp. 2892–2900. doi: 10.1109/CVPR.2015.7298881.
29. [29] Z. Zhou, M. M. R. 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, pp. 3–11, Springer, 2018. doi: 10.1007/978-3-030-00928-1_1.
30. [30] Y. Peng, M. Sonka and D. Z. Chen, “U-Net v2: rethinking the skip connections of U-Net for medical image segmentation,” arXiv preprint arXiv:2311.17791, 2023.
31. [31] J. Chen, Y. Lu, Q. Yu, X. Luo, E. Adeli, Y. Wang and Y. Zhou, “TransUNet: transformers make strong encoders for medical image segmentation,” arXiv preprint arXiv:2102.04306, 2021.
32. [32] S. He, R. Bao, P. E. Grant and Y. Ou, “U-Netmer: U-Net meets transformer for medical image segmentation,” arXiv preprint arXiv:2304.01401, 2023.
33. [33] O. Oktay et al., “Attention U-Net: learning where to look for the pancreas,” arXiv preprint arXiv:1804.03999, 2018.
34. [34] Y. Zhang, C. Wang and W. Ji, “Attention-gated networks for medical image segmentation: a survey of methods and applications,” Computers in Biology and Medicine, vol. 149, p. 105939, 2022. doi: 10.1016/j.compbiomed.2022.105939