SHIP DETECTION AND CLASSIFICATION USING YOLOV8 THROUGH TRANSFER LEARNING APPROACH

Abstract

 In urgent situations such as maritime disasters rapid vessel detection and tracking are vital for coordinating search and rescue, as well as for enabling autonomous navigation and collision avoidance systems. This study evaluates the use of YOLOv8m, a high‑performance one‑stage deep‑learning detector, to locate ships in profile imagery and categorize them by type. By adopting a transfer‑learning strategy, we fine‑tune a pre‑trained YOLOv8m model on a diverse compilation of open‑source ship photographs, adapting its general object‑recognition capabilities to this specialized maritime task. We report on the model’s performance measured by detection recall, precision, and classification accuracy demonstrating its suitability for real‑time ship identification in varied operational scenarios. And testing with various classes for ship identification.