基于偶然不确定性的单人姿态估计模型测试时增强算法  

A Testing-Time-Augmentation Algorithm for Single Human Pose Estimation Based on Aleatoric Uncertainty

作　　者：李杰[1,2,3] 亓波 张建林[2,3] LI Jie;QI Bo;ZHANG Jianlin(Key Lab.of Optical Beam Control Engineering,Chinese Academy of Sciences,Chengdu 610209,CHN;Institute of Optics and Electronics Technology,Chinese Academy of Sciences,Chengdu 610209,CHN;University of Chinese Academy of Sciences,Beijing 100049,CHN)

机构地区：[1]中国科学院光束控制重点实验室,成都610209 [2]中国科学院光电技术研究所,成都610209 [3]中国科学院大学,北京100049

出　　处：《半导体光电》2022年第5期968-973,共6页Semiconductor Optoelectronics

摘　　要：针对现有单人姿态估计网络结果缺乏可靠性评估和鲁棒性保障等问题,提出了一种基于偶然不确定性的测试时增强方法。该方法首先利用随机并行的数据增强和模型推理得到多样化输出,随后通过计算该输出的偶然不确定性得到其可靠性评估,最后根据可靠性将该输出及其不确定性进行加权融合以得到更准确鲁棒的结果及其评估。在MPII数据集上的实验表明,该算法可即插即用地应用于任意现有单人姿态估计网络,从而得到更精确鲁棒的结果及其不确定性评估。Aiming at the problem of lacking reliability evaluation and robustness guarantee in existing single-person pose estimation networks’results,a testing-time-augmentation(TTA)algorithm based on aleatoric uncertainty was proposed.In this TTA algorithm,diverse outputs were firstly obtained by stochastic parallel data augmentation and model inference.Then,the reliability evaluations of those outputs are acquired by calculating their aleatoric uncertainty.Finally,those outputs and their uncertainty were fused according to the reliabilities to obtain a more accurate and robust result as well as its evaluation.Experiments on the MPII dataset show that this algorithm can be applied to any existing single-person pose estimation network in a plug-and-play manner,leading to a more precise and robust result with its uncertainty evaluation.

关 键 词：单人姿态估计 关键点检测 偶然不确定性 测试时增强 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]