IMPACT OF DATA AUGMENTATION ON TRAINING COMPUTER VISION MODEL FOR SHIPSʼ ASPECT ANGLE DETECTION
https://doi.org/10.33815/2313-4763.2025.2.31.052-063
Abstract
Abstract. The study investigates the impact of targeted data augmentation strategies on the performance of a deep learning-based computer vision model designed to determine a shipsʼ aspect angle in maritime scenes. The work addresses the challenge posed by limited and highly imbalanced datasets that are typical for maritime imaging, where variations in illumination, weather, and occlusion significantly affect recognition accuracy. The aim of the research is to evaluate which augmentation techniques contribute to improved robustness of YOLOv8n-based orientation classification without distorting the physical characteristics of marine scenes. The methodology combines controlled experiments with color, geometric and spatially local augmentation, applied to an extended dataset containing both annotated examples and negative samples. The scientific novelty lies in identifying the augmentation techniques that deliver the highest computer vision model performance according to the key metrics mAP50, mAP50-95, precision, and recall, while ensuring that these techniques enhance the performance of a model specifically designed for operation in maritime environments. The results demonstrate that color and geometry-preserving augmentations, such as translation and scaling, yield measurable improvements, while perspective distortions severely degrade performance due to the violation of spatial realism. The obtained results have practical significance for autonomous navigation systems, contributing to improved recognition accuracy under real operatinal conditions. The conclusions highlight that augmentation strategies must be carefully selected for tasks involving orientation-sensitive objects.
References
2. Yang, D., Solihin, M. I., Ardiyanto, I., Zhao, Y., Li, W., Cai, B., & Chen, C. (2024). A streamlined approach for intelligent ship object detection using EL-YOLO algorithm. Scientific Reports, 14, 15254. https://doi.org/10.1038/s41598-024-64225-y.
3. Langenkämper, D., van Kevelaer, R., & Nattkemper, T. W. (2019). Strategies for Tackling the Class Imbalance Problem in Marine Image Classification. In Z. Zhang, D. Suter, Y. Tian, A. Branzan Albu, N. Sidère, & H. J. Escalante (Eds.), Pattern Recognition and Information Forensics: ICPR International Workshops and Challenges (pp. 26–36). Springer. https://doi.org/10.1007/978-3-030-05792-3_3.
4. McGuinness, K., & O'Gara, S. (2019). Comparing data augmentation strategies for deep image classification. In J. Courtney, C. Deegan, & P. Leamy (Eds.), IMVIP 2019 Irish Machine Vision & Image Processing Conference proceedings. Irish Pattern Recognition & Classification Society. https://doras.dcu.ie/23666/.
5. Wang, H., Deng, Z., Jiang, C., Ren, J., & Zhang, S. (2022). A review of deep learning techniques for maritime object detection from optical remote sensing images. Sensors, 22(14), 5383. https://doi.org/10.3390/s22145383.
6. Li, Y., Zhang, X., Zhang, Y., Wang, R., & Li, J. (2024). YOLO-SEA: An enhanced detection framework for multi-scale maritime targets in complex sea states and adverse weather. Sensors, 24(1), 243. https://doi.org/10.3390/s24010243.
7. Wang, Z., Wang, P., Liu, K., Wang, P., Fu, Y., Lu, C. T., Aggarwal, C. C., Pei, J., & Zhou, Y. (2024). A comprehensive survey on data augmentation. arXiv preprint arXiv:2405.09591.
8. Shorten, C., & Khoshgoftaar, T. M. (2021). A survey on image data augmentation for deep learning. Journal of Big Data, 6(1), 60. https://doi.org/10.1186/s40537-019-0197-0.
9. Xu, M., Yoon, S., Fuentes, A., & Park, D. S. (2023). A comprehensive survey of image augmentation techniques for deep learning. Pattern Recognition, 137, 109347. https://doi.org/10.1016/j.patcog.2023.109347.
10. Tan, M., Langenkämper, D., & Nattkemper, T. W. (2022). The impact of data augmentations on deep learning-based marine object classification in benthic image transects. Sensors, 22(14), 5383. https://doi.org/10.3390/s22145383.
11. Cauli, N., & Reforgiato Recupero, D. (2022). Survey on videos data augmentation for deep learning models. Future Internet, 14(3), 93. https://doi.org/10.3390/fi14030093.
12. Pashenko O. L., Pipchenko O. D. (2025). Rozrobka modeli shtuchnoho zoru na bazi alhorytmu YOLO dlia vyznachennia rakursu sudna. Sudnovodinnia (Shipping & Navigation). 38.
13. Cauli, N., & Reforgiato Recupero, D. (2022). Survey on Videos Data Augmentation for Deep Learning Models. Future Internet, 14(3), 93. https://doi.org/10.3390/fi14030093.
14. Mumuni, A., & Mumuni, F. (2022). Data augmentation: A comprehensive survey of modern approaches. Array, 16, 100258. https://doi.org/10.1016/j.array.2022.100258.
15. Kumar, T., Brennan, R., Mileo, A., & Bendechache, M. (2024). Image Data Augmentation Approaches: A Comprehensive Survey and Future Directions. IEEE Access, 12, 187536-187571. https://doi.org/10.48550/arXiv.2301.02830.
