New Findings on Robotics from Bannari Amman Institute of Technology Summarized ( Implementation of Hand-object Pose Estimation Using Ssd and Yolov5 Model for Obj ect Grasping By Scara Robot)
New Findings on Robotics from Bannari Amman Institute of Technology Summarized ( Implementation of Hand-object Pose Estimation Using Ssd and Yolov5 Model for Obj ect Grasping By Scara Robot)
一位新闻记者兼机器人与机器学习的新闻编辑每日新闻-机器人的最新数据在一份新的报告中呈现。根据sRx新编辑在印度Sathyamangalam的新闻报道，研究表明：“在人-机器人协作任务中，应用先进的深度学习方法进行手-物体姿态估计是一种必不可少的方法。在遮挡、临界光照、光照等多种情况下，从二维图像中提取手-物体的位置和方向仍然是一个关键问题。”以及显著区域检测和模糊图像。我们的新闻记者引用了Bannari Amman Institute of Technology的研究:“本文提出的方法使用增强的Mob ileNetV3和单镜头检测(SSD)和YOLOv5,在保证提高精度的同时,不影响手物体姿态和方位检测的延迟,克服了计算量大、延迟和准确性高等局限性。”MobileNetV3采用了网络结构搜索和NetAdapt算法,实现了参数整定的网络搜索和多尺度特征提取的自适应学习和锚盒偏移调整,压缩激励块减少了模型的计算量和延迟,并采用了硬交换函数和特征金字塔网络来防止超分辨率。在对Mobi leNetV3与其前身和YOLOV5进行对比分析的基础上,得到的准确率分别为92.8%和89.7%,召回率分别为93.1%和90.2%,mAP值分别为93.3%和89.2%。
Abstract
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News – Fresh data on Robotics are presented i n a new report. According to news reporting out of Sathyamangalam, India, by New sRx editors, research stated, “Enforcement of advanced deep learning methods in hand-object pose estimation is an imperative method for grasping the objects saf ely during the human-robot collaborative tasks. The position and orientation of a hand-object from a two-dimensional image is still a crucial problem under vari ous circumstances like occlusion, critical lighting, and salient region detectio n and blur images.” Our news journalists obtained a quote from the research from the Bannari Amman I nstitute of Technology, “In this paper, the proposed method uses an enhanced Mob ileNetV3 with single shot detection (SSD) and YOLOv5 to ensure the improvement i n accuracy and without compromising the latency in the detection of hand-object pose and its orientation. To overcome the limitations of higher computation cost , latency and accuracy, the Network Architecture Search and NetAdapt Algorithm i s used in MobileNetV3 that perform the network search for parameter tuning and a daptive learning for multiscale feature extraction and anchor box offset adjustm ent due to auto-variance of weight in the level of each layers. The squeeze-and- excitation block reduces the computation and latency of the model. Hard-swish ac tivation function and feature pyramid networks are used to prevent over fitting the data and stabilizing the training. Based on the comparative analysis of Mobi leNetV3 with its predecessor and YOLOV5 are carried out, the obtained results ar e 92.8% and 89.7% of precision value, recall value o f 93.1% and 90.2%, mAP value of 93.3% a nd 89.2%, respectively.”
Key words
Sathyamangalam/India/Asia/Emerging Te chnologies/Machine Learning/Robot/Robotics/Bannari Amman Institute of Techno logy
NETL
NSTL