baiduyunguangjia_cfg_A900527E-5BA6-4d22-8E96-E40D5C6EDF61.cfg

The data set contains three postures of live pigs (standing, lying on their side, and lying on their stomach).The Posture_4520_my_yolo_dataset contains 4520 images. According to the ratio of 8:1:1, the training set (3616), the verification set (452), the test set (452).Posture_7220_my_yolo_dataset, this dataset was obtained by amplification based on 4520 annotated images.The images in the Posture_7220_my_yolo_dataset were relatively good, and the model accuracy was difficult to be further improved. Subsequent experiments suggested that a small number of blocked and crowded samples should be selected for annotation on the basis of the 4520 dataset to further improve the model.

本数据集涵盖生猪的三类姿态：站立、侧卧与趴卧。其中Posture_4520_my_yolo_dataset 共包含4520张图像，按照8:1:1的比例划分为训练集（3616张）、验证集（452张）与测试集（452张）。Posture_7220_my_yolo_dataset 是基于4520张已标注图像经数据扩增得到的数据集，其内部图像整体质量较佳，难以通过该数据集进一步提升模型精度。后续实验表明，在4520数据集的基础上选取少量被遮挡且场景拥挤的样本进行标注，可进一步优化模型性能。