姿态估计中常用的评价指标

MPJPE/MPVPE: measures the mean per joint/vertex position error by Euclidean distance (mm) between the estimated and ground-truth coordinates.
PA-MPJPE/MPVPE: is a modification of MPJPE/MPVPE with Procrustes analysis, ignoring global variation. For conciseness, this metric is abbreviated as PJ/PV.
Acc: captures the acceleration of 2D/3D joint landmarks in pixel/$s^2$ or mm/$s^2$ to reflect temporal coherence.
AUC: means the area under the curve of PCK (percentage of correct keypoints) vs. error thresholds of n ∼ 50mm (for 3D measurement) or 0 ∼ 30 pixel (for 2D measurement). According to - , n = 0 or 20.
F-Score: is the harmonic mean between recall and precision between two meshes w.r.t. a specific distance threshold. F@5/F@15 corresponds to a threshold of 5mm/15mm.

MPJPE/MPVPE: 通过估计的和真实坐标之间的欧几里德距离（mm）测量每个关节/顶点位置误差的平均值。
PA-MPJPE/MPVPE: 是对MPJPE/MPVPE的修改，采用Procrustes分析，忽略了全局变化。为简明起见，该指标缩写为PJ/PV。
Acc: 以像素/$s^2$或mm/$s ^2$捕捉2D/3D关节界标的加速度，以反映时间一致性。
AUC: 是指PCK曲线下的面积，PCK曲线是指将正确预测的关键点数（正确预测的关键点数占总关键点数的比例）作为横坐标，错误的关键点到真实位置的欧几里得距离（在三维空间中的距离，或者在二维图像中的像素距离）作为纵坐标，绘制出的曲线。在计算AUC时，需要对PCK曲线的横坐标进行限制，即只考虑误差阈值在一定范围内的预测结果。对于三维测量，误差阈值的范围为n ∼ 50mm，而对于二维测量，误差阈值的范围为0 ∼ 30像素。在误差阈值范围内，PCK曲线的下方面积即为AUC的值。误差阈值的范围可以是n = 0或n = 20。

*F-Score: 是两个网格之间的召回和精度之间的调和平均值*w.r.t.*一个特定的距离阈值。F@5/F@15对应于5mm/15mm的阈值。