摘要
奶牛脸部及关键点检测可以帮助牛场构建识别系统以及对牛脸进行姿态估计,帮助牛场智能化升级。基于深度学习的奶牛脸部及关键点检测模型需要数据集进行训练,以提高模型的准确性和鲁棒性。因此,在自然条件拍摄的奶牛脸部数据集,对于奶牛脸部及关键点检测模型训练,实现牛脸识别系统搭建以及牛脸姿态估计至关重要。本数据集包含了不同角度、不同光照、不同数量等复杂环境下的奶牛脸部的图像,共2538张图像。使用Labelme软件对奶牛脸部以及左右眼睛中心、鼻子中间、左右嘴角5个关键点进行标注,得到2538个json标注文件。并将标注的COCO(json)格式文件转为YOLO(txt)格式文件,把图片以及对应的txt格式标签按照4:1的比例划分训练集和测试集,用于训练YOLOv5-Pose、YOLOv7-Pose、YOLOv8-Pose等关键点检测模型。实验表明,本数据集在奶牛脸部及关键点检测模型上表现良好,为奶牛脸部及关键点检测等方向上的研究和应用提供了有价值的图像数据资源。
Cow face and keypoint detection can assist farms in building recognition systems and estimating cow face poses,facilitating the intelligent upgrade of farms.A deep learning-based model for cow face and keypoint detection requires a dataset for training to improve the model’s accuracy and robustness.Therefore,a dataset of cow faces captured under natural conditions is crucial for training cow face and keypoint detection models,which are essential for building cow face recognition systems and estimating cow face poses.This dataset contains 2,538 images of cow faces captured under various lighting conditions,including different angles,lighting,and cow numbers.Labelme software was used to annotate the cow faces and five key points:the centers of the left and right eyes,the middle of the nose,and the left and right corners of the mouth,resulting in 2,538 json annotation files.The annotated COCO(json)format files were then converted to YOLO(txt)format files.The images and corresponding txt format labels were divided into training and test sets in a 4:1 ratio.These datasets were used to train keypoint detection models such as YOLOv5-Pose,YOLOv7-Pose,and YOLOv8-Pose.Experiments have shown that this dataset performs well in cow face and keypoint detection models,providing valuable image data resources for research and applications in cow face and keypoint detection.
作者
侯现坤
黄小平
黄飞
豆子豪
郑寰宇
王晨洋
冯涛
刘梦艺
HOU Xiankun;HUANG Xiaoping;HUANG Fei;DOU Zihao;ZHENG Huanyu;WANG Chenyang;FENG Tao;LIU Mengyi(National Engineering Research Center for Agro-Ecological Big Data Analysis&Application,School of Internet,Anhui University,Hefei 230039,P.R.China)
基金
安徽省科技厅自然科学基金项目(2308085MC103)
安徽省教育厅自然科学重点项目(KJ2021A0024)
安徽省高等学校自然科学基金(2023AH050082)。
关键词
奶牛
深度学习
牛脸检测
关键点检测
cow
deep learning
cow face detection
keypoint detection
