The rehabilitation treatment helps to heal minor sports and occupational
injuries. In a traditional rehabilitation process, a therapist will assign
certain actions to a patient to perform in between hospital visits, and it will
rely on the patient to remember actions correctly and the schedule to perform
them. Unfortunately, many patients forget to perform actions or fail to recall
actions in detail. As a consequence, the rehabilitation treatment is hampered
or, in the worst case, the patient may suffer from additional injury caused by
performing incorrect actions. To resolve these issues, we propose a HRNet-based
rehabilitation monitoring system, which can remind a patient when to perform
the actions and display the actions for the patient to follow via the patient's
smartphone. In addition, it helps the therapist to monitor the progress of the
rehabilitation for the patient. Our system consists of an iOS app and several
components at the server side. The app is in charge of displaying and
collecting action videos. The server computes the similarity score between the
therapist's actions and the patient's in the videos to keep track of the number
of repetitions of each action. Theses stats will be shown to both of the
patient and therapist. The extensive experiments show that the F1-Score of the
similarity calculation is as high as 0.9 and the soft accuracy of the number of
repetitions is higher than 90%.

本研究提出了基于 HRNet 的康复监测系统，该系统可通过患者的智能手机提醒其何时执行康复动作，并展示动作的视频，同时帮助治疗师监测患者的康复进展，本系统的相似度计算的 F1 得分高达 0.9，动作次数的软准确性高达 90％以上。

基于 HRNet 的康复监测系统

A HRNet-based Rehabilitation Monitoring System

The mainstream researche in deep metric learning can be divided into two
genres: proxy-based and pair-based methods. Proxy-based methods have attracted
extensive attention due to the lower training complexity and fast network
convergence. However, these methods have limitations as the poxy optimization
is done by network, which makes it challenging for the proxy to accurately
represent the feature distrubtion of the real class of data. In this paper, we
propose a Calibrate Proxy (CP) structure, which uses the real sample
information to improve the similarity calculation in proxy-based loss and
introduces a calibration loss to constraint the proxy optimization towards the
center of the class features. At the same time, we set a small number of
proxies for each class to alleviate the impact of intra-class differences on
retrieval performance. The effectiveness of our method is evaluated by
extensive experiments on three public datasets and multiple synthetic
label-noise datasets. The results show that our approach can effectively
improve the performance of commonly used proxy-based losses on both regular and
noisy datasets.

本文提出了一种 Calibrate Proxy (CP) 结构，它使用真实样本信息提高了代理损失中的相似度计算，并引入校准损失来约束代理优化向类特征中心。 通过多个实验验证了我们方法在三个公共数据集和多个合成噪音标签数据集上的有效性，表明我们的方法可以有效地改善常用代理损失在正常和噪声数据集上的表现。