Fitness-AQA

该数据集包含三种运动项目的视频：深蹲、杠铃划船和肩上推举，并由专业教练针对运动错误进行了标注。这些视频是在健身房真实环境下录制的，展现了在现实条件下自然发生的运动错误，标注内容覆盖了从细微到严重的各种错误。该数据集的任务是对锻炼姿势的评估以及错误检测。

This dataset includes videos of three exercise types: back squats, barbell rows, and overhead presses, which are annotated by professional coaches for exercise form errors. These videos were recorded in real gym environments, showcasing naturally occurring movement errors under realistic conditions, and the annotations cover errors ranging from subtle to severe. The tasks supported by this dataset are exercise posture assessment and error detection.

搜集汇总

背景与挑战

背景概述

Fitness-AQA是一个用于细粒度健身动作质量评估的数据集，包含三种健身动作（深蹲、过头推举和杠铃划船），旨在通过AI技术检测和纠正健身动作中的错误姿势。该数据集特别针对真实健身房场景中的挑战，如摄像头角度、服装风格和器材遮挡，提供了自监督表示学习方法。