首页> 外国专利> METHOD, APPARATUS, ELECTRONIC DEVICE AND, READABLE STORAGE MEDIUM AND PROGRAM FOR CONSTRUCTING KEY-POINT LEARNING MODEL

METHOD, APPARATUS, ELECTRONIC DEVICE AND, READABLE STORAGE MEDIUM AND PROGRAM FOR CONSTRUCTING KEY-POINT LEARNING MODEL

机译:方法,装置,电子设备和,可读存储介质和用于构建关键点学习模型的程序

摘要

An embodiment of the present invention discloses a key point learning model building method, apparatus, electronic device and computer readable storage medium, but relates to the fields of image processing, image labeling, artificial intelligence, and machine learning technology. A specific embodiment of the method includes: acquiring labeling data in which the human body key point is already labeled and non-labeling data in which the human body key point is not labeled; Next, training the labeling data according to a supervised training method through an initial predictive model and an initial discriminator to obtain a first predictive model and a first discriminator; then, training the unlabeled data according to an unsupervised training method through the first predictive model and the first discriminator to obtain a second predictive model and a second discriminator; Finally, building a key point learning model according to the second predictive model and the second discriminator. The above embodiment reduces the conditions and difficulty of building a key point learning model by fully utilizing the non-labeling data existing in large quantities.
机译:本发明的一个实施例公开了一种关键点学习模型构建的方法,装置,电子设备和计算机可读存储介质,但涉及图像处理,图像标记,人工智能,和机器学习技术的领域。该方法的一个具体实施例包括:获取在其中人体关键点已标记的,并且其中所述人体关键点未标记的非标记数据的标记数据;接着,根据通过初始预测模型和初始鉴别器以获得第一预测模型和一个第一鉴别有监督训练方法训练标记数据;然后,根据通过第一预测模型和所述第一鉴别器以获得第二预测模型和第二鉴别器无监督的训练方法训练未标记数据;最后,根据第二预测模型和第二鉴别建设的一个关键点学习模式。上述实施例减少的条件和充分利用现有的大量非标记数据构建一个关键点学习模型的难度。

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号