Journal of the Korean Society of Marine Environment & Safety (해양환경안전학회지)

The Korean Society of Marine Environment and safety (해양환경안전학회)
권호 표지
DOI QR코드

DOI QR Code

Comparative Study of Fish Detection and Classification Performance Using the YOLOv8-Seg Model

YOLOv8-Seg 모델을 이용한 어류 탐지 및 분류 성능 비교연구

  • Sang-Yeup Jin (GeoSystem Research Corp.) ;
  • Heung-Bae Choi (GeoSystem Research Corp.) ;
  • Myeong-Soo Han (GeoSystem Research Corp.) ;
  • Hyo-tae Lee (Korea Fisheries Resources Agency) ;
  • Young-Tae Son (GeoSystem Research Corp.)
  • 진상엽 ((주)지오시스템리서치) ;
  • 최흥배 ((주)지오시스템리서치) ;
  • 한명수 ((주)지오시스템리서치) ;
  • 이효태 (한국수산자원공단 자원회복실) ;
  • 손영태 ((주)지오시스템리서치)
  • Received : 2024.03.06
  • Accepted : 2024.04.26
  • Published : 2024.04.30
Download PDF
⟨ Previous Next ⟩

Abstract

The sustainable management and enhancement of marine resources are becoming increasingly important issues worldwide. This study was conducted in response to these challenges, focusing on the development and performance comparison of fish detection and classification models as part of a deep learning-based technique for assessing the effectiveness of marine resource enhancement projects initiated by the Korea Fisheries Resources Agency. The aim was to select the optimal model by training various sizes of YOLOv8-Seg models on a fish image dataset and comparing each performance metric. The dataset used for model construction consisted of 36,749 images and label files of 12 different species of fish, with data diversity enhanced through the application of augmentation techniques during training. When training and validating five different YOLOv8-Seg models under identical conditions, the medium-sized YOLOv8m-Seg model showed high learning efficiency and excellent detection and classification performance, with the shortest training time of 13 h and 12 min, an of 0.933, and an inference speed of 9.6 ms. Considering the balance between each performance metric, this was deemed the most efficient model for meeting real-time processing requirements. The use of such real-time fish detection and classification models could enable effective surveys of marine resource enhancement projects, suggesting the need for ongoing performance improvements and further research.

수산자원의 지속 가능한 관리와 증대는 전 세계적으로 중요한 이슈로 부상하고 있으며, 본 연구는 이에 대응하는 한국수산자원공단의 수산자원 현존량 추정을 위한 딥러닝 기반 수산자원 증대사업 효과조사 기법 개발을 위해 구성 기술 중 하나인 어류 탐지 및 분류 모델 구축과 성능 비교를 수행하였다. 다양한 크기의 YOLOv8-Seg 모델에 어류 이미지 데이터셋을 학습한 후 각 성능평가 지표를 비교 분석하여 적용 가능한 최적의 모델을 선정하고자 하였다. 모델 구축에 사용된 자료는 총 12종의 어류로 이루어진 36,749장의 이미지와 라벨 파일로 이루어지며, 학습에는 증강을 적용하여 데이터의 다양성을 증가시켰다. 동일한 환경 및 조건에서 총 다섯 개의 YOLOv8-Seg 모델을 학습 및 검증한 결과 중간 크기의 YOLOv8m-Seg 모델이 가장 짧은 13시간 12분의 학습 시간과 mAP50:95 0.933, 추론 속도 9.6 ms로 높은 학습 효율성과 우수한 탐지 및 분류 성능을 보였으며, 각 지표 간의 균형을 고려할 때 실시간 처리 요구사항을 충족하는 가장 효율적인 모델로 평가되었다. 이와 같은 실시간 어류 탐지 및 분류 모델을 활용하여 효율적인 수산자원 증대사업의 효과조사가 가능할 것으로 보이며, 지속적인 성능 개선 및 추가적인 연구가 필요할 것으로 사료된다.

Keywords

Acknowledgement

이 논문은 한국수산자원공단의 '딥러닝 기반 수산자원 증대사업 효과조사 기법 개발(2023)' 사업의 지원을 받아 수행되었으며, 한국지능정보사회진흥원(NIA) AIHub의 '어류 개체 촬영 영상' 자료를 활용하였습니다.

References

  1. Akgul, T., N. Calik, and B. U. Toreyin(2020), Bulanik Sualti Goruntulerinde Derin Ogrenme Tabanli Balik Tespiti Deep Learning-Based Fish Detection in Turbid Underwater Images.
  2. Bai, R., M. Wang, Z. Zhang, J. Lu, and F. Shen(2023), Automated Construction Site Monitoring Based on Improved YOLOv8-seg Instance Segmentation Algorithm. IEEE Access, 11, 139082-139096. https://doi.org/10.1109/ACCESS.2023.3340895.
  3. Bolya, D., C. Zhou, F. Xiao, and Y. J. Lee(2019), YOLACT: Real-Time Instance Segmentation. 2019 IEEE/CVF International Conference on Computer Vision (ICCV), 9156-9165. https://doi.org/10.1109/ICCV.2019.00925.
  4. Chen, L. -C., G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille(2017), DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs (arXiv:1606.00915). arXiv. http://arxiv.org/abs/1606.00915.
  5. Climent-Perez, P., A. Galan-Cuenca, N. E. Garcia-d'Urso, M. Saval-Calvo, J. Azorin-Lopez, and A. Fuster-Guillo(2024), Simultaneous, vision-based fish instance segmentation, species classification and size regression. PeerJ Computer Science, 10, e1770. https://doi.org/10.7717/peerj-cs.1770.
  6. Dumitriu, A., F. Tatui, F. Miron, R. T. Ionescu, and R. Timofte(2023), Rip Current Segmentation: A Novel Benchmark and YOLOv8 Baseline Results. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 1261-1271. https://doi.org/10.1109/CVPRW59228.2023.00133.
  7. Fabic, J. N., I. E. Turla, J. A. Capacillo, L. T. David, and P. C. Naval(2013), Fish population estimation and species classification from underwater video sequences using blob counting and shape analysis. 2013 IEEE International Underwater Technology Symposium (UT), pp. 1-6. https://doi.org/10.1109/UT.2013.6519876.
  8. He, K., G. Gkioxari, P. Dollar, and R. Girshick(2017), Mask r-cnn. In Proceedings of the IEEE international conference on computer vision, pp. 2961-2969.
  9. Li, X., M. Shang, H. Qin, and L. Chen(2015), Fast accurate fish detection and recognition of underwater images with Fast R-CNN. OCEANS 2015 - MTS/IEEE Washington, 1-5. https://doi.org/10.23919/OCEANS.2015.7404464.
  10. Long, J., E. Shelhamer, and T. Darrell(2015), Fully Convolutional Networks for Semantic Segmentation.
  11. Padilla, R., S. L. Netto, and E. A. B. Da Silva(2020), A Survey on Performance Metrics for Object-Detection Algorithms. 2020 International Conference on Systems, Signals and Image Processing (IWSSIP), pp. 237-242. https://doi.org/10.1109/IWSSIP48289.2020.9145130.
  12. Ravanbakhsh, M., M. R. Shortis, F. Shafait, A. Mian, E. S. Harvey, and J. W. Seager(2015), Automated Fish Detection in Underwater Images Using Shape Based Level Sets. The Photogrammetric Record, 30(149), 46-62. https://doi.org/10.1111/phor.12091.
  13. Tian, G., D. Li, W. Li, L. Zhang, H. Zhang, and Q. Duan (2021), A detection method of the turned white belly fish based on improved SSD. Journal of Physics: Conference Series, 1856(1), 012035. https://doi.org/10.1088/1742-6596/1856/1/012035.
  14. Varatharasan, V., H. -S. Shin, A. Tsourdos, and N. Colosimo (2019), Improving Learning Effectiveness For Object Detection and Classification in Cluttered Backgrounds. 2019 Workshop on Research, Education and Development of Unmanned Aerial Systems (RED UAS), 78-85. https://doi.org/10.1109/REDUAS47371.2019.8999695.
  15. Wang, C. -Y., A. Bochkovskiy, and H. -Y. M. Liao(2022), YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors (arXiv:2207.02696). arXiv. http://arxiv.org/abs/2207.02696.
  16. Wang, X., R. Zhang, T. Kong, L. Li, and C. Shen(2020), SOLOv2: Dynamic and Fast Instance Segmentation.
  17. Xu, W. and S. Matzner(2018), Underwater Fish Detection Using Deep Learning for Water Power Applications. 2018 International Conference on Computational Science and Computational Intelligence (CSCI), pp. 313-318. https://doi.org/10.1109/CSCI46756.2018.00067.