@article{10.11648/j.ijema.20251306.13,
  author = {Vinod Vadde and Takur Sai Karthik Chalamalasetty and Paka Shalini and Wisam Bukaita},
  title = {Automated Detection and Classification of Underwater Species Using YOLOv8 for Real-time Marine Ecosystem Monitoring},
  journal = {International Journal of Environmental Monitoring and Analysis},
  volume = {13},
  number = {6},
  pages = {314-327},
  doi = {10.11648/j.ijema.20251306.13},
  url = {https://doi.org/10.11648/j.ijema.20251306.13},
  eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.ijema.20251306.13},
  abstract = {Effective monitoring of marine biodiversity is essential for understanding ecosystem health, detecting species population changes, and mitigating the impacts of environmental degradation. Traditional underwater observation techniques, such as diver-based surveys and manual video analysis, are labor-intensive, time-consuming, and prone to human error. Consequently, there is an increasing need for automated, data-driven methods capable of performing real-time detection and analysis of aquatic species under diverse environmental conditions. This study introduces a deep learning framework based on the YOLOv8 architecture for automated detection, classification, and segmentation of underwater species. A curated dataset containing seven representative classes fish, jellyfish, starfish, shark, puffin, penguin, and crown-of-thorns starfish is used for model training and evaluation. Data preprocessing techniques, including image enhancement, resizing, and normalization, were applied to address underwater imaging challenges such as low contrast, noise, and color distortion. The model was trained using transfer learning and data augmentation to improve robustness and generalization under varying light and turbidity conditions. The experimental results demonstrate that the proposed YOLOv8 framework achieves Precision of 80.82%, Recall of 69.35%, mAP@0.5 of 76.86%, and mAP@0.5–0.95 of 47.57% in object detection tasks. The segmentation module further attained 85.48% accuracy, enabling precise delineation of species boundaries for morphological assessment. These outcomes highlight YOLOv8’s superior ability to generalize across diverse underwater environments compared to conventional convolutional neural network (CNN)–based approaches. Overall, this research presents a scalable and efficient deep learning solution for real-time underwater species monitoring. The integration of detection and segmentation capabilities enables accurate, fine-grained analysis that can enhance marine conservation, ecological assessment, and automated biodiversity mapping. The proposed YOLOv8-based framework represents a significant step toward the deployment of intelligent visual systems in marine ecosystem monitoring and environmental sustainability applications.},
 year = {2025}
}

Copy | Download

TY  - JOUR
T1  - Automated Detection and Classification of Underwater Species Using YOLOv8 for Real-time Marine Ecosystem Monitoring
AU  - Vinod Vadde
AU  - Takur Sai Karthik Chalamalasetty
AU  - Paka Shalini
AU  - Wisam Bukaita
Y1  - 2025/12/09
PY  - 2025
N1  - https://doi.org/10.11648/j.ijema.20251306.13
DO  - 10.11648/j.ijema.20251306.13
T2  - International Journal of Environmental Monitoring and Analysis
JF  - International Journal of Environmental Monitoring and Analysis
JO  - International Journal of Environmental Monitoring and Analysis
SP  - 314
EP  - 327
PB  - Science Publishing Group
SN  - 2328-7667
UR  - https://doi.org/10.11648/j.ijema.20251306.13
AB  - Effective monitoring of marine biodiversity is essential for understanding ecosystem health, detecting species population changes, and mitigating the impacts of environmental degradation. Traditional underwater observation techniques, such as diver-based surveys and manual video analysis, are labor-intensive, time-consuming, and prone to human error. Consequently, there is an increasing need for automated, data-driven methods capable of performing real-time detection and analysis of aquatic species under diverse environmental conditions. This study introduces a deep learning framework based on the YOLOv8 architecture for automated detection, classification, and segmentation of underwater species. A curated dataset containing seven representative classes fish, jellyfish, starfish, shark, puffin, penguin, and crown-of-thorns starfish is used for model training and evaluation. Data preprocessing techniques, including image enhancement, resizing, and normalization, were applied to address underwater imaging challenges such as low contrast, noise, and color distortion. The model was trained using transfer learning and data augmentation to improve robustness and generalization under varying light and turbidity conditions. The experimental results demonstrate that the proposed YOLOv8 framework achieves Precision of 80.82%, Recall of 69.35%, mAP@0.5 of 76.86%, and mAP@0.5–0.95 of 47.57% in object detection tasks. The segmentation module further attained 85.48% accuracy, enabling precise delineation of species boundaries for morphological assessment. These outcomes highlight YOLOv8’s superior ability to generalize across diverse underwater environments compared to conventional convolutional neural network (CNN)–based approaches. Overall, this research presents a scalable and efficient deep learning solution for real-time underwater species monitoring. The integration of detection and segmentation capabilities enables accurate, fine-grained analysis that can enhance marine conservation, ecological assessment, and automated biodiversity mapping. The proposed YOLOv8-based framework represents a significant step toward the deployment of intelligent visual systems in marine ecosystem monitoring and environmental sustainability applications.
VL  - 13
IS  - 6
ER  - 

Copy | Download