Sports-102, Yoga-82, Yoga-107, Dance-12

本研究提出了四个用于细粒度人体姿势识别的数据集，重点关注体育、瑜伽和舞蹈姿势。这些数据集包括102个类别的Sports-102、82个类别的Yoga-82、107个类别的Yoga-107和12个类别的Dance-12。这些数据集用于训练和评估提出的SYD-Net模型，该模型集成了基于补丁的注意力机制，以提高姿势识别的性能。This study proposes four datasets for fine-grained human pose recognition, focusing on sports, yoga and dance poses. Specifically, they are Sports-102 with 102 categories, Yoga-82 with 82 categories, Yoga-107 with 107 categories, and Dance-12 with 12 categories. These datasets are utilized to train and evaluate the proposed SYD-Net model, which integrates a patch-based attention mechanism to improve the performance of pose recognition.

搜集汇总

背景与挑战

背景概述

Sports-102、Yoga-82、Yoga-107和Dance-12是四个用于细粒度人体姿势识别的数据集，分别涵盖体育、瑜伽和舞蹈姿势类别，旨在支持SYD-Net模型的训练与评估，该模型通过基于补丁的注意力机制优化识别性能。