plSSN : 0374-8111 | elSSN : 2287-8815
plSSN : 0374-8111 | elSSN : 2287-8815
plSSN : 0374-8111elSSN : 2287-8815
Identifying Major Clinical Signs of Scuticociliatosis in Olive Flounder Paralichthys olivaceus Using Deep-Learning-Based Object Detection
Ri Zheng†·Helin Yin†·조미영1·한현자1·김아란1·이남실1·정준범2·구영현*·임재웅1*
세종대학교 인공지능데이터사이언스학과, 1국립수산과학원 병리연구과, 2제주대학교 수산생명의학과
The olive flounder Paralichthys olivaceus is a key aquaculture species in South Korea, accounting for over half of the total farmed fish production in the country. However, frequent disease outbreaks, particularly scuticociliatosis, have economically harmed the aquaculture industry. Conventional approaches to diagnosing scuticociliatosis, such as visual inspection and molecular testing, are subjective, expensive, and unsuitable for field deployment. We developed a two-stage image-based diagnostic pipeline for identifying the major clinical signs of scuticociliatosis using deep learning-based object detection. the first, the olive flounder is localized within the image. Second, six external clinical signs (ulceration, corrosion, erosion, erythema, hernia, and distension) were detected. We evaluated the performance of four object detection models: Faster R-CNN, DETR, YOLO v8, and Co-DETR. The Co-DETR model was superior, with a precision of 94.0% and recall of 97.6%. A prototype web application was developed for practical use in field environments. This study demonstrates the feasibility of applying deep learning-based visual diagnosis in aquaculture and provides a foundation for the development of automated disease detection tools for various fish species.
Computer vision, Deep learning, Object detection, Olive flounder, Scuticociliatosis