Yoga-82
收藏资源简介:
Yoga-82是由大阪大学和印度理工学院甘地讷格尔分校合作创建的大型瑜伽姿势识别数据集,包含82个类别,总计超过28,400张图像。该数据集通过Bing搜索引擎从网络下载,并手工清理和标注,每张图片包含一个或多个执行相同瑜伽姿势的人。数据集的创建旨在通过细粒度的层次化姿势分类,解决现有数据集在姿势多样性、物体遮挡和视角方面的不足。Yoga-82通过提供基于身体配置的层次化标签,包括身体位置、身体位置的变化和实际姿势名称,来解决复杂姿势的精细标注问题。该数据集的应用领域包括视觉监控、行为分析、辅助生活和智能驾驶辅助系统等。
Yoga-82 is a large-scale yoga pose recognition dataset co-created by Osaka University and Indian Institute of Technology Gandhinagar. It contains 82 categories and a total of over 28,400 images. This dataset is downloaded from the web via Bing Search Engine, followed by manual cleaning and annotation. Each image includes one or more people performing the same yoga pose. The dataset is developed to address the shortcomings of existing datasets in terms of pose diversity, object occlusion and viewing angles through fine-grained hierarchical pose classification. Yoga-82 solves the problem of fine annotation for complex poses by providing hierarchical labels based on body configuration, which cover body positions, variations of body positions and actual pose names. The application fields of this dataset include visual surveillance, behavior analysis, assisted living, intelligent driving assistance systems and other related areas.




