广东海洋大学高体鰤幼苗外形监测图像标注数据集

数据集来源于广东海洋大学养殖基地，通过学校采购的霸勒思水下摄像机、SJCAM运动相机及普联水面摄像机多角度采集原始视频，覆盖水下/水面双视角及早/中/晚多时段光线条件。数据处理包含三个核心环节：视频抽帧与筛选（使用FFmpeg按固定间隔抽取原始帧，经质量筛选保留合格图像）；图像数据增强：针对低光照与夜景图像，分别采用对比度增强与边缘增强技术扩展数据规模； 图像智能标注：结合手工标注与YOLOv8预训练模型自动标注，经人工核验确保标注精度（误差≤5像素），最终形成1062张图像并按7:2:1比例划分为训练集、验证集和测试集。数据主要应用于支持幼苗实时识别计数、投喂决策优化及健康状态监测；为多光源/多角度算法验证、幼苗行为研究提供标准数据；通过量化幼苗尺寸变化及光照适应性分析，建立生产评估基准；作为水产信息化课程与科研全流程训练的实践素材。

This dataset is collected from the breeding base of Guangdong Ocean University. Raw videos were captured from multiple perspectives using school-procured Baleisi underwater cameras, SJCAM action cameras and TP-Link water surface cameras, covering both underwater and water surface viewpoints as well as varying lighting conditions across early, mid and late periods of a day. The data processing includes three core steps: 1. Video frame extraction and screening: Original frames were extracted at fixed intervals using FFmpeg, and qualified images were retained after quality filtering; 2. Image data augmentation: For low-light and night-scene images, contrast enhancement and edge enhancement techniques were respectively adopted to expand the dataset scale; 3. Intelligent image annotation: Integrating manual annotation and automatic annotation via the pre-trained YOLOv8 model, manual verification was conducted to guarantee annotation accuracy (error ≤ 5 pixels). Finally, a total of 1062 qualified images were generated, which were split into training, validation and test sets at a ratio of 7:2:1. This dataset is primarily applied to support real-time aquatic seedling recognition and counting, feeding decision optimization and health status monitoring; it provides standard benchmark data for multi-light source/multi-angle algorithm verification and seedling behavior research; it establishes production evaluation benchmarks by quantifying seedling size variations and light adaptability analysis; and it serves as practical training materials for the entire process of aquaculture informatization courses and scientific research.

搜集汇总

背景与挑战

背景概述

该数据集是广东海洋大学高体鰤幼苗外形监测图像标注数据集，包含1062张经过多角度采集和智能标注的图像，主要用于幼苗识别计数、投喂决策优化及健康状态监测等水产养殖智能管理。