广东海洋大学草鱼异常行为图像标注数据集

本数据集是图像标注的草鱼异常行为数据集。数据集核心在于系统化捕捉并标注了草鱼在实验室封闭养殖环境中（涵盖水下/水面双视角及早中晚多时段光线变化）表现的四类异常行为。数据集通过视频抽帧、图像增强及半自动标注（人工预标注结合YOLOv8改进模型）流程构建，包含对应的XML标注文件。该数据集可用于水产养殖智能监控系统的实时异常行为识别预警、AI管理平台及水下机器人巡检，弥补了草鱼集约化养殖中精细化异常行为监测的数据不足，为智能水产养殖科研机构与企业提供了高效的解决方案。

This is an image-annotated abnormal behavior dataset for grass carp. The core of this dataset lies in the systematic capture and annotation of four types of abnormal behaviors exhibited by grass carp in a closed laboratory aquaculture environment, which covers dual perspectives of underwater and water surface as well as light condition changes across morning, noon and evening. This dataset is constructed through workflows including video frame extraction, image enhancement, and semi-automatic annotation (manual pre-annotation combined with an improved YOLOv8 model), and includes corresponding XML annotation files. This dataset can be applied to real-time abnormal behavior recognition and early warning for intelligent aquaculture monitoring systems, AI management platforms, and underwater robot inspection tasks. It fills the gap of insufficient data for fine-grained abnormal behavior monitoring in intensive grass carp aquaculture, providing efficient solutions for scientific research institutions and enterprises engaged in intelligent aquaculture.

搜集汇总

背景与挑战

背景概述

广东海洋大学草鱼异常行为图像标注数据集是一个包含3214张图像的数据集，涵盖了草鱼在实验室封闭养殖环境中的四类异常行为。数据集通过视频抽帧、图像增强及半自动标注流程构建，适用于水产养殖智能监控系统的实时异常行为识别预警、AI管理平台及水下机器人巡检。