Sheep Facial Expression Primary Data-set

The sheep facial expression primary dataset has been cropped of the collection sheep imagesfrom ImageNet, NADIS, Pixabay, Flickr, and Gettyimages with high resolutions. It consists of 1500 plus sheep face images which has two subset normal and abnormal sheep faces. The abnormality of sheep can be identify by ears, eyes and mouth of the sheep. If the ears pinna visible than have no pain else for less visible and not visible have low and high pain respectively. Similarly for eyes and mouth will depends upon V and U shape. Here we uploaded a sample subset and complete dataset. This dataset has been used in our coming research article. We have aim to collect more images which will be helpful to find pain in sheep through deep learning. As deep learning needs big data to complete a desired task from scratch, while currently we used transfer learning to find abnormality of sheep from there faces with deep learning.本绵羊面部表情基础数据集的高清图像均采集自ImageNet、NADIS、Pixabay、Flickr及Gettyimages平台，并经过裁剪预处理。该数据集共包含1500余张绵羊面部图像，分为正常绵羊面部与异常绵羊面部两个子集。绵羊面部的异常状态可通过其耳部、眼部与口部特征进行判别：当耳郭（ear pinna）可见时，绵羊无疼痛症状；耳郭可见度较低时，绵羊存在轻度疼痛；耳郭完全不可见时，则代表绵羊存在重度疼痛。同理，眼部与口部的异常状态需根据其呈现的V形与U形特征进行判别。本次我们上传了该数据集的样本子集与完整数据集。该数据集已应用于我们即将发表的研究论文中。我们计划进一步采集更多图像数据，以助力通过深度学习技术实现绵羊疼痛状态的识别。由于深度学习模型从零开始完成指定任务需要大规模数据支撑，因此本次研究我们采用了迁移学习（transfer learning）方法，通过绵羊面部图像实现其异常状态的判别。