Fish species identification plays a vital role in fisheries management, biodiversity conservation, aquaculture, and marine ecosystem monitoring. Accurate and automated detection of fish species from underwater images can significantly reduce manual effort while improving monitoring efficiency. Traditional image classification approaches often struggle with varying underwater conditions such as poor illumination, occlusion, complex backgrounds, and differences in fish orientation and size. To address these challenges, this study proposes a YOLOv8-based deep learning framework for real-time fish species detection and classification. A publicly available fish image dataset was annotated using bounding boxes corresponding to different fish species and used to train the model with GPU acceleration for improved computational efficiency. The proposed framework performs simultaneous localization and species classification, enabling accurate identification of multiple fish within a single image. Experimental evaluation was conducted using standard object detection metrics, and the proposed model achieved a mean Average Precision (mAP@0.5) of 81.91%, demonstrating strong performance in detecting and classifying fish species across diverse underwater environments. The developed system can support intelligent fisheries management, marine biodiversity assessment, ecological research, and automated monitoring applications by providing fast, reliable, and accurate fish species detection with minimal human intervention.

Sri Sanjana K , Samyukthaa N, vishalini P, Umesh M S, Ms Vidhyabharathi T,"YOLOv8-Assisted Underwater Fish Species Detection From Underwater Images" International Journal of Advanced Multidisciplinary Research and Educational Development, V2(4): Page(517-533) July - August 2026. ISSN: 3107-6513. www.ijamred.com. Published by Scientific and Academic Research Publishing.