基于改进YOLO-Pose轻量模型的多人姿态估计

YOLO-Pose作为人体姿态估计算法模型,在精度和速度上有着不错的表现,但其在复杂和有遮挡的场景下存在误检率较大的问题,并且模型的复杂度仍然有优化的空间.针对这几个问题,通过选取Slim-neck模块和Res2Net模块,重新设计其特征融合层,减少其计算量和参数量,提高特征提取能力,在提升精度的同时,使模型轻量化;引入EIoU损失函数,加快边框检测的收敛速度,并提高定位的准确性.在压缩的OC_Human数据集上进行测试,改进后的模型与YOLO-Pose相比,P值、mAP@0.5和mAP@.5:95分别提高了10.6,3.1和2.9个百分点.此外,参数量和计算量也分别减少了16.7%和19.3%,在精度和轻量化方面均有所提升,为其应用在资源有限的边缘计算设备提供了可能性.

基于改进YOLOv8-pose的分心驾驶检测与识别

针对现有的分心驾驶检测算法存在检测率低、检测速率慢等问题,本文构建了一种基于改进YOLOv8-pose的分心驾驶检测识别模型YOLOv8-EFM。首先,通过使用EfficientViT更换YOLOv8-pose的主干网络,结合CNN和VIT之间的互补性,提升了检测的准确率;其次,使用FasterBlock模块替换C2f中的Bottleneck模块,增加了检测速率并减小模型参数;最后在SPPF后加入了轻量级的MLCA注意力模块,在模型大小和准确性之间取得了良好的平衡。实验结果表明,本文所构建的YOLOv8-EFM模型,mAP50可以达到98.9%,模型大小只有9.7 M,该方法不仅可以识别出具体分心行为,还可以检测上半身的人体骨架,可以有效应用在驾驶员分心驾驶的检测场景中。

基于改进YOLOv8n-Pose的轨道作业人员跨轨安全动作识别

针对轨道作业人员跨轨安全动作监督方法存在效率低、漏检率高等问题,引入改进的人体姿态估计算法YOLOv8n-Pose对跨轨安全动作进行识别和监督。对YOLOv8n-Pose算法改进方法为在网络中添加注意力机制并轻量化网络结构,并改进网络的bbox损失函数和关键点损失函数,以提高网络的识别精度和速度。使用高斯滤波和ColorJitter算法对自制数据集增强。在训练前使用遗传算法对训练超参数进行自适应调整,在训练时使用迁移学习和知识蒸馏方法,提高网络训练速度、识别精度和泛化能力。将训练好的模型对轨道现场作业人员图像进行检测,可成功识别出作业人员姿态并根据关键点位置信息识别安全动作,人体关键点识别精确度为94.3%,推理速度为238.1 fps,验证模型改进研究取得了有益效果,提高了模型识别精度、识别速度和鲁棒性。

基于Yolopose的挖掘机检测与工作状态识别

针对光缆、高压油气管道等地下基础设施周边容易受到挖掘机的野蛮入侵问题.本文提出了一种结合Yolopose和多层感知机的挖掘机检测与工作状态判别方法.首先,设计了基于Yolopose的挖掘机6点姿势的提取网络Yolopose-ex;其次,利用Yolopose-ex模型提取视频中挖掘机工作姿态的变化信息,构建了挖掘机的工作状态特征向量(MSV);最后,利用深度学习算法多层感知机(multilayer perceptron,MLP)分析了视频中的挖掘机的工作状态.实验结果表明,所提出的方法克服了复杂背景难以识别的问题,对挖掘机工作状态识别准确率达到了96.6%,具有较高的推理速度和泛化能力.

基于MediaPipe Pose的人体动作识别方法研究

针对已有人体动作识别方法存在识别效率低、检测速度慢等问题,提出了基于MediaPipe Pose算法的人体动作识别方法。具体内容:将摄像头实时采集数据输入到检测网络以获取人体33个关键点的坐标信息,然后通过关键点的空间位置组合来确定人体动作类别;采用COCO数据集格式标定动作类别,并且对动作标签进行onehot编码,训练人体动作识别模型;利用单目RGB摄像头对8类动作进行实验验证。结果表明,基于MediaPipe Pose算法的人体动作识别方法其帧率达到30帧/s,识别精确率为96.67%,能够实时、准确地识别人体动作。

基于YOLO-Pose的城市街景小目标行人姿态估计算法

现有的姿态估计算法在城市街景中对小目标行人的检测效果不佳。针对该问题,提出一种基于YOLO-Pose的小目标行人姿态估计算法YOLO-Pose-CBAM。通过引入CBAM注意力机制模块,在不增加过多计算量的前提下,增强网络聚焦小目标行人区域的能力,提升算法对小目标行人的敏感度,同时在主干网络中使用4个不同尺寸的检测头,丰富算法对图片中不同大小行人的检测手段;在骨干网络和颈部之间架设2条跨层级联通道,提升浅层网络与深层网络之间的特征融合能力,进一步增强信息交流,降低小目标行人漏检率;引入SIoU重新定义边界框回归的定位损失函数,加快训练的收敛速度,提高检测精度;采用k-means++算法代替k-means算法对数据集中标注的锚框进行聚类,避免聚类中心初始化时导致的局部最优解问题,从而选择出更适合检测小目标行人的锚框。对比实验结果表明,在小目标行人Wider Keypoints数据集上,所提算法相较于YOLO-Pose和YOLOv7-Pose在平均精度上分别提升了4.6和6.5个百分比。

基于改进YOLOv8n-pose和三维点云分析的蒙古马体尺自动测量方法

[目的/意义]准确高效地获取马匹体尺信息是马产业现代化进程中的关键环节。传统的人工测量方法耗时长、工作量大,且会对马匹造成一定应激反应。因此,实现准确且高效的体尺参数自动测量对于制定蒙古马早期育种计划至关重要。[方法]选择Azure Kinect深度相机获取蒙古马双侧RGB-D数据,以YOLOv8n-pose为基础,通过在C2f模块中引入可变形卷积(Deformable Convolution v2, DCNv2),同时添加洗牌注意力机制(Shuffle Attention, SA)模块和优化损失函数(SCYLLA-IoU Loss, SIoU)的方法,利用余弦退火法动态调整学习率,提出一种名为DSS-YOLO (DCNv2-SA-SIoU-YOLO)的模型用于蒙古马体尺关键点的检测。其次,将RGB图中的二维关键点坐标与深度图中对应深度值相结合,得到关键点三维坐标,并实现蒙古马点云信息的转换。利用直通滤波、随机抽样一致性(Random Sample Consensus, RANSAC)、统计离群值滤波、主成分分析法(Principal Component Analysis, PCA)完成点云处理与分析。最终根据关键点坐标自动计算体高、体斜长、臀高、胸围和臀围5项体尺参数。[结果和讨论] DSS-YOLO的平均关键点检测精度为92.5%;d_(DSS)为7.2个像素;参数量和运算量分别仅为3.48 M和9.1 G。体尺参数自动测量结果与人工测量值相比,各项体尺参数的整体平均绝对误差为3.77 cm;平均相对误差为2.29%。[结论]研究结果可为蒙古马运动性能相关遗传参数的确定提供技术支撑。

Lightweight Multi-Resolution Network for Human Pose Estimation

Human pose estimation aims to localize the body joints from image or video data.With the development of deeplearning,pose estimation has become a hot research topic in the field of computer vision.In recent years,humanpose estimation has achieved great success in multiple fields such as animation and sports.However,to obtainaccurate positioning results,existing methods may suffer from large model sizes,a high number of parameters,and increased complexity,leading to high computing costs.In this paper,we propose a new lightweight featureencoder to construct a high-resolution network that reduces the number of parameters and lowers the computingcost.We also introduced a semantic enhancement module that improves global feature extraction and networkperformance by combining channel and spatial dimensions.Furthermore,we propose a dense connected spatialpyramid pooling module to compensate for the decrease in image resolution and information loss in the network.Finally,ourmethod effectively reduces the number of parameters and complexitywhile ensuring high performance.Extensive experiments show that our method achieves a competitive performance while dramatically reducing thenumber of parameters,and operational complexity.Specifically,our method can obtain 89.9%AP score on MPIIVAL,while the number of parameters and the complexity of operations were reduced by 41%and 36%,respectively.

Movement Function Assessment Based on Human Pose Estimation from Multi-View

Human pose estimation is a basic and critical task in the field of computer vision that involves determining the position(or spatial coordinates)of the joints of the human body in a given image or video.It is widely used in motion analysis,medical evaluation,and behavior monitoring.In this paper,the authors propose a method for multi-view human pose estimation.Two image sensors were placed orthogonally with respect to each other to capture the pose of the subject as they moved,and this yielded accurate and comprehensive results of three-dimensional(3D)motion reconstruction that helped capture their multi-directional poses.Following this,we propose a method based on 3D pose estimation to assess the similarity of the features of motion of patients with motor dysfunction by comparing differences between their range of motion and that of normal subjects.We converted these differences into Fugl–Meyer assessment(FMA)scores in order to quantify them.Finally,we implemented the proposed method in the Unity framework,and built a Virtual Reality platform that provides users with human–computer interaction to make the task more enjoyable for them and ensure their active participation in the assessment process.The goal is to provide a suitable means of assessing movement disorders without requiring the immediate supervision of a physician.

Investigation of Inside-Out Tracking Methods for Six Degrees of Freedom Pose Estimation of a Smartphone in Augmented Reality

Six degrees of freedom(6DoF)input interfaces are essential formanipulating virtual objects through translation or rotation in three-dimensional(3D)space.A traditional outside-in tracking controller requires the installation of expensive hardware in advance.While inside-out tracking controllers have been proposed,they often suffer from limitations such as interaction limited to the tracking range of the sensor(e.g.,a sensor on the head-mounted display(HMD))or the need for pose value modification to function as an input interface(e.g.,a sensor on the controller).This study investigates 6DoF pose estimation methods without restricting the tracking range,using a smartphone as a controller in augmented reality(AR)environments.Our approach involves proposing methods for estimating the initial pose of the controller and correcting the pose using an inside-out tracking approach.In addition,seven pose estimation algorithms were presented as candidates depending on the tracking range of the device sensor,the tracking method(e.g.,marker recognition,visual-inertial odometry(VIO)),and whether modification of the initial pose is necessary.Through two experiments(discrete and continuous data),the performance of the algorithms was evaluated.The results demonstrate enhanced final pose accuracy achieved by correcting the initial pose.Furthermore,the importance of selecting the tracking algorithm based on the tracking range of the devices and the actual input value of the 3D interaction was emphasized.

Abnormal Action Recognition with Lightweight Pose Estimation Network in Electric Power Training Scene

Electric power training is essential for ensuring the safety and reliability of the system.In this study,we introduce a novel Abnormal Action Recognition(AAR)system that utilizes a Lightweight Pose Estimation Network(LPEN)to efficiently and effectively detect abnormal fall-down and trespass incidents in electric power training scenarios.The LPEN network,comprising three stages—MobileNet,Initial Stage,and Refinement Stage—is employed to swiftly extract image features,detect human key points,and refine them for accurate analysis.Subsequently,a Pose-aware Action Analysis Module(PAAM)captures the positional coordinates of human skeletal points in each frame.Finally,an Abnormal Action Inference Module(AAIM)evaluates whether abnormal fall-down or unauthorized trespass behavior is occurring.For fall-down recognition,three criteria—falling speed,main angles of skeletal points,and the person’s bounding box—are considered.To identify unauthorized trespass,emphasis is placed on the position of the ankles.Extensive experiments validate the effectiveness and efficiency of the proposed system in ensuring the safety and reliability of electric power training.

DAUNet: Detail-Aware U-Shaped Network for 2D Human Pose Estimation

Human pose estimation is a critical research area in the field of computer vision,playing a significant role in applications such as human-computer interaction,behavior analysis,and action recognition.In this paper,we propose a U-shaped keypoint detection network(DAUNet)based on an improved ResNet subsampling structure and spatial grouping mechanism.This network addresses key challenges in traditional methods,such as information loss,large network redundancy,and insufficient sensitivity to low-resolution features.DAUNet is composed of three main components.First,we introduce an improved BottleNeck block that employs partial convolution and strip pooling to reduce computational load and mitigate feature loss.Second,after upsampling,the network eliminates redundant features,improving the overall efficiency.Finally,a lightweight spatial grouping attention mechanism is applied to enhance low-resolution semantic features within the feature map,allowing for better restoration of the original image size and higher accuracy.Experimental results demonstrate that DAUNet achieves superior accuracy compared to most existing keypoint detection models,with a mean PCKh@0.5 score of 91.6%on the MPII dataset and an AP of 76.1%on the COCO dataset.Moreover,real-world experiments further validate the robustness and generalizability of DAUNet for detecting human bodies in unknown environments,highlighting its potential for broader applications.

基于改进YOLOv5s-pose的多人人体姿态估计

为了提高多人人体姿态检测的准确率,本研究采用YOLOv5s模型用于多人人体姿态检测并对模型进行改进。首先,引入坐标注意力(Coordinate Attention)模块改进骨干网络,将注意力资源分配给关键区域,降低复杂环境中的背景干扰,增强模型对多人目标的精准定位能力。其次,使用双向特征金字塔网络改进YOLOv5s的特征融合网络,增强网络的信息表达能力。实验结果表明:在多人人体姿态MS COCO2017验证集上,经改进的YOLOv5s算法的检测平均精度高达61.9%,相比原始YOLOv5s网络,平均精度提升了1.5%。由此可见,改进后的网络能更加精准、有效地检测多人人体姿态。

基于OpenPose的晕眩警告设计

文章采用了Open Pose识别人体摔倒姿态特征,提取人体关键性的关节节点。鉴于摔倒和站立的人体姿态有着明显的区别,仅仅认识到在地上就判定摔倒是不准确的,还需要结合站立动作特征。因此,识别两种动作特征后,可发现可能晕眩并及时通知保卫室,从而有效降低因晕眩可能导致无人察觉的安全隐患,确保人身安全。

Survey on depth and RGB image-based 3D hand shape and pose estimation

The field of vision-based human hand three-dimensional(3D)shape and pose estimation has attracted significant attention recently owing to its key role in various applications,such as natural human computer interactions.With the availability of large-scale annotated hand datasets and the rapid developments of deep neural networks(DNNs),numerous DNN-based data-driven methods have been proposed for accurate and rapid hand shape and pose estimation.Nonetheless,the existence of complicated hand articulation,depth and scale ambiguities,occlusions,and finger similarity remain challenging.In this study,we present a comprehensive survey of state-of-the-art 3D hand shape and pose estimation approaches using RGB-D cameras.Related RGB-D cameras,hand datasets,and a performance analysis are also discussed to provide a holistic view of recent achievements.We also discuss the research potential of this rapidly growing field.

改进PoseConv3D模型在建筑工人临边不安全行为识别中的应用

为准确识别建筑工人的常见危险动作行为,降低高处坠落事故的发生率,提出了一种基于改进PoseConv3D的建筑工人临边不安全行为识别方法。该方法在PoseConv3D行为识别模型的基础上,采用高分辨率网络(High-Resolution Network,HRNet)姿态估计模型,用于捕捉工人骨骼关键点的模态信息,并在行为特征提取网络中引入三维卷积注意力机制(3D Convolutional Block Attention Module,3D-CBAM),以强化关键特征的自适应学习和分配能力。此外,依据风险等级选取跨越护栏、抛物、吸烟、站姿倚靠护栏、坐姿倚靠护栏和打电话六类不安全行为进行数据集构建。在自行构建的不安全行为数据集上的试验结果表明,该方法对六类临边危险行为动作的Top-1准确率可达95.3%,具有良好的识别精度和泛化性,能够准确识别建筑临边场景下工人的危险行为动作。

基于改进YOLO-Pose的复杂环境下拖拉机驾驶员关键点检测

为解决农田复杂作业环境下拖拉机驾驶员因光照、背景及遮挡造成的关键点漏检、误检等难识别问题,该研究提出了一种基于改进YOLO-Pose的复杂环境下驾驶员关键点检测方法。首先,在主干网络的顶层C3模块中嵌入Swin Transformer编码器,提高遮挡状况下关键点的检测效率。其次,采用高效层聚合网络RepGFPN作为颈部网络,通过融合高层语义信息和低层空间信息,增强多尺度检测能力,同时在颈部网络采用金字塔卷积替换标准3×3卷积,在减少模型参数量的同时有效地捕获不同层级的特征信息。最后,嵌入坐标注意力机制优化关键点解耦头,增强预测过程对关键点空间位置的敏感程度。试验结果表明,改进后算法mAP0.5(目标关键点相似度Loks阈值取0.5时平均精度均值)为89.59%,mAP0.5:0.95(目标关键点相似度Loks阈值取0.5,0.55,…,0.95时的平均精度均值)为62.58%,相比于基线模型分别提高了4.24和4.15个百分点,单张图像平均检测时间为21.9 ms,与当前主流关键点检测网络Hourglass、HRNet-W32及DEKR相比,mAP0.5分别提升了7.94、5.27、2.66个百分点,模型大小分别减少了257.5、8.2、9.3 M。改进后的关键点检测算法具有较高的检测精度和推理速度,可为拖拉机驾驶员的异常行为识别和状态监测提供技术支持。

基于改进YOLO v8-Pose的红熟期草莓识别和果柄检测

针对高架栽培模式下的大棚草莓,借鉴人体姿态检测算法,建立了改进YOLO v8-Pose模型对红熟期草莓进行识别与果柄关键点检测。通过对比YOLO v5-Pose、YOLO v7-Pose、YOLO v8-Pose模型,确定使用YOLO v8-Pose模型作为对红熟期草莓识别与关键点预测的模型。以YOLO v8-Pose为基础,对其网络结构添加Slim-neck模块与CBAM注意力机制模块,提高模型对小目标物体的特征提取能力,以适应草莓数据集的特点。改进YOLO v8-Pose能够有效检测红熟期草莓并准确标记出果柄关键点,P、R、mAP-kp分别为98.14%、94.54%、97.91%,比YOLO v8-Pose分别提高5.41、5.31、8.29个百分点。模型内存占用量为22 MB,比YOLO v8-Pose的占用量小6 MB。此外,针对果园非结构化的特征,探究了光线、遮挡与拍摄角度对模型预测的影响。对比改进前后的模型在复杂环境下对红熟期草莓的识别与果柄预测情况,改进YOLO v8-Pose在受遮挡、光线和角度影响情况下的mAPkp分别为94.52%、95.48%、94.63%,较YOLO v8-Pose分别提高8.9、10.75、5.17个百分点。改进YOLO v8-Pose可在保证网络模型精度的同时对遮挡、光线和拍摄角度等影响均具有较好的鲁棒性,能够实现对复杂环境下红熟期草莓识别与果柄关键点预测。

基于YOLOPose的人体姿态估计轻量级网络

为解决人体姿态估计模型在提升预测精度时参数量和计算量增多导致模型运行效率低下的问题,在YOLOPose模型基础上设计出一种轻量级人体姿态估计网络MWE-YOLOPose。选择轻量级MobileNetV3网络重新构建骨干网络,保持特征丰富性同时加快特征提取速度;调整特征融合层通道数并添加ECA注意力机制进行跨通道交互,实现模型轻量化与准确度的平衡;引用WIOUV2损失函数降低几何因素的惩罚,增强模型的鲁棒性和泛化能力。实验结果显示,在OC_Human数据集上,改进后模型对比原始YOLOPose模型,在保持一定准确度的情况下,模型参数量和计算量分别降低86.8%和71.2%,有效降低了模型运算复杂度。

Image-Based Brachytherapy in Cervical Cancer: Review and Experiences in Faculty of Medicine, Chiang Mai University

Cervical cancer is the one of the most common cancer in female patients inThailand. Radiotherapy has the role for the treatment of cervical cancer by postoperative, radical and palliative treatments. For radical radiotherapy, the combination of external beam radiation therapy and brachytherapy will be used to increase the tumor dose to curative goal. With the new development of medical images (Computed tomography (CT), Magnetic Resonance Imaging (MRI) or Ultrasonography (US)), the treatment with brachytherapy will be developed from point-based to volume-based concepts. Many studies reported the benefit of image-based brachytherapy over conventional brachytherapy and clinical benefit of using image-based brachytherapy in the treatment of cervical cancer.