Article Detail

Ahmet İlisulu* , Cemal Turan, Yakup Kutlu

Keywords

Segmentation YOLO Computer Vision Object Detection

Doi : 10.71350/jere.2026182

Abstract

Automated fish detection and identification are critical for aquaculture and marine ecology research. However, the complexity of underwater environments and the difficulty of pixel-level labeling limit the performance of computer vision models. This study aims to detect and visualize the outer boundaries of fish using closed polygon coordinates with the YOLO architecture. A standardized, clean dataset of 2700 images containing 13 scientific fish species was created. The model was trained and generated, and predictions for determining the boundaries of fish, converted to a closed polygon labeling format, were rendered as closed edge contours on the original images. According to the validation set results, the model demonstrated high performance in the mask prediction task, achieving an F1 score of 0.934 and mAP50-95 values ​​of 0.854. Furthermore, all fish outer boundaries were successfully drawn on the image using the established automated visualization pipeline.

References

1. Alsuwaylimi, A. A. (2024). Enhanced YOLOv8-Seg Instance Segmentation for Real-Time Submerged Debris Detection. IEEE Access, 12, 114825-114835.
2. Automated Mask Generation for Efficient YOLO-Based Instance Segmentation in Marine Environments for Fish Detection. (2024). Marine Vision and Technology Journal, 8(2), 142-155.
3. Chowdhury, S., Kutlu, Y., Goongoon, N. (2026). Comparative Evaluation of YOLO-Based Pretrained Models for Multi-Class Vegetable Detection. Tethys Environmental Science, 3(1), 16-30, doi : 10.71350/tethysjournal.2
4. İlisulu, A., Kutlu, Y. (2025). YOLO-Based Deep Learning Approach for Turkish Sign Language Recognition. Journal of Natural and Engineering Research, 1(1), 38-44, doi : 10.5281/zenodo.18664511
5. Kutlu, Y., Altan, G., İşçimen, B., Doğdu, S. A., & Turan, C. (2017a). Recognition of species of Triglidae Family using Deep Learning. Journal of the Black Sea/Mediterranean Environment, 23(1), 56-65.
6. Kutlu, Y., İşçimen, B., & Turan, C. (2017b). Multi-Stage Fish Classification System Using Morphometry. Fresenius Environmental Bulletin, 26(3), 1910-1916.
7. Kutlu, Y., Turan, C.(2018). An Intelligent Software for Measurements of Biological Materials: BioMorph. Natural and Engineering Sciences (Yayın No: 4709858)
8. Liu, Y., Fan, H., Wang, S., Huang, Y., Luo, N., Ren, H., & Xu, H. (2026). Tongue Image Segmentation Using YOLOv11-based Object Detection and Mobile Segment Anything Model. In 2026 6th International Conference on Neural Networks, Information and Communication Engineering (NNICE) (pp. 411-417). IEEE.
9. Qin, H., Mei, Y., Yu, H., Xu, L., Chai, Y., He, Q., & Chen, Y. (2026). Improved YOLO-InceptionV3 network for shrimp detection in recirculation aquaculture system. Information Processing in Agriculture (Baskıda).
10. Suchodolak, M. A. R., Fraidenraich, G., de Lima, E. R., & de Oliveira, J. R. (2026). A Semi-Automatic Labeling Approach for Image Segmentation Using YOLOv11. IEEE Access, 14, 49908-49927.
11. Wang, J., Cheng, Z., Lin, M., Yang, R., & Huang, Q. (2025). FishKP-YOLOv11: An Automatic Estimation Model for Fish Size and Mass in Complex Underwater Environments. Animals, 15(19), 2862.
12. Zhang, H., Yang, X., Zhang, R., Gao, N., Wang, N., Zhang, Z., & Zhang, C. (2026). Detection of dense fish schools in sonar imagery with a novel YOLOv11-SAS model. Ecological Informatics, 94, 103646.