Providing ground truth supervision to train visual models has been a bottleneck over the years, exacerbated by domain shifts which degenerate the performance of such models. This was the case when visual tasks relied on handcrafted features and shallow machine learning and, despite its unprecedented performance gains, the problem remains open within the deep learning paradigm due to its data-hungry nature. Best performing deep vision-based object detectors are trained in a supervised manner by relying on human-labeled bounding boxes which localize class instances (i.e.objects) within the training images.Thus, object detection is one of such tasks for which human labeling is a major bottleneck. In this paper, we assess co-training as a semi-supervised learning method for self-labeling objects in unlabeled images, so reducing the human-labeling effort for developing deep object detectors. Our study pays special attention to a scenario involving domain shift; in particular, when we have automatically generated virtual-world images with object bounding boxes and we have real-world images which are unlabeled. Moreover, we are particularly interested in using co-training for deep object detection in the context of driver assistance systems and/or self-driving vehicles. Thus, using well-established datasets and protocols for object detection in these application contexts, we will show how co-training is a paradigm worth to pursue for alleviating object labeling, working both alone and together with task-agnostic domain adaptation.
AcousticFusion: Fusing Sound Source Localization to Visual SLAM in Dynamic Environments
Shenzhen Institute of Artificial Intelligence and Robotics for Society（深圳人工智能与机器人研究院）；the Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））；Westlake University（西湖大学）；Westlake Institute for Advanced Study（西湖高等研究院）；The University of Edinburgh（爱丁堡大学）；The Alan Turing Institute（艾伦·图灵研究所）
CD-TWINSAFE: A ROS-enabled Digital Twin for Scene Understanding and Safety Emerging V2I Technology
C-DRiVeS Lab: Cognitive Driving Research in Vehicular Systems, Cairo, Egypt（认知驾驶研究与车辆系统实验室，埃及开罗）；Computer Science and Engineering Department - Faculty of Media Engineering and Technology（计算机科学与工程系-媒体工程与技术学院）；German University in Cairo, Egypt（埃及开罗德国大学）
Benchmarking Edge Computing Devices for Grape Bunches and Trunks Detection using Accelerated Object Detection Single Shot MultiBox Deep Learning Models
INESC TEC – Instituto de Engenharia, Tecnologia e Ciencia（INESC TEC——工程与科技研究所）；Faculty of Engineering, University of Porto（波尔图大学工程学院）；Nottingham Trent University（诺丁汉特伦特大学）；Khalifa University of Science（哈利法科学大学）
DeepRadiologyNet: Radiologist Level Pathology Detection in CT Head Images
DeepRadiology, INC.（DeepRadiology 公司）；University of California, Los Angeles（加州大学洛杉矶分校）；University of California, San Diego（加州大学圣迭戈分校）；University of Oxford（牛津大学）
Deep Learning Based Defect Detection for Solder Joints on Industrial X-Ray Circuit Board Images
National ASIC Research Center（国家专用集成电路研究中心）；Southeast University（东南大学）；Singapore University of Technology and Design（新加坡科技设计大学）；Peking University（北京大学）；Keysight Company（是德科技公司）
On-board classification of underwater images using hybrid classical-quantum CNN based method
1 Department of Physics, Mahindra University, Hyderabad, 500043, Telangana, India；2 Department of Computer Science, Mahindra University, Hyderabad, 500043, Telangana, India；3 Department of Mechanical \& Aerospace Engineering, Mahindra University, Hyderabad, 500043, Telangana, India
Deep Learning for Real Time Satellite Pose Estimation on Low Power Edge TPU
Thales Alenia Space Italia（泰雷兹阿莱尼亚航天意大利公司）
不合作空间驻留物体的位姿估计是近距离操作中实现自主性的关键资产。在这种情况下，单目相机是一种有价值的解决方案，因为它们的系统要求低。然而，相关联的图像处理算法对于实时的板上实现来说在计算上太昂贵，或者不够准确。在本文中，我们提出了一个姿态估计软件，利用神经网络架构，可以扩展到不同的精度-延迟权衡。我们设计了与边缘张量处理单元兼容的管道，以展示低功耗机器学习加速器如何在太空中实现人工智能开发。测试的神经网络的基准航天器姿态估计数据集，并特意开发的Cosmo真实感数据集，它描绘了一个COSMO-SkyMed卫星在各种随机姿势和可操纵的太阳能电池板的方向。我们架构的最轻版本在两个数据集上都实现了最先进的准确性，但网络复杂性只有一小部分，在Coral Dev Board Mini上以每秒7.7帧的速度运行，仅消耗2.2W。
Image Processing Lab, Universidad de Valencia（图像处理实验室，瓦伦西亚大学）；Center for Biomaterials and Tissue Engineering Universitat Politecnica de Valencia（生物材料与组织工程中心，瓦伦西亚理工大学）
ESRPCB: an Edge guided Super-Resolution model and Ensemble learning for tiny Printed Circuit Board Defect detection
Faculty of Electronics and Telecommunications, University of Engineering and Technology, Vietnam National University（电子与电信学院，工程与技术大学，越南国家大学）；School of Information Science, Japan Advanced Institute of Science and Technology（信息科学学院，日本先进科学和技术研究院）；Faculty of Mechatronics, SMAE, Hanoi University of Industry（机械学院，SMAE，河内工业大学）
COVID-19爆发后，自动检测人们是否戴口罩以降低一线工作人员的风险变得非常重要。此外，在本地处理用户数据是解决隐私和网络带宽问题的好方法。在本文中，我们提出了一个轻量级模型，用于检测特定区域的人是否佩戴口罩，该模型也可以部署在Coral Dev Board上，Coral Dev Board是一种包含Google Edge TPU的商用开发板。我们的方法结合了基于MobileNetV 2加SSD的对象检测网络和仅限整数硬件的量化方案。因此，Edge TPU中更轻的模型具有更低的延迟，更适合实时的【形】执行，同时保持与浮点器件相当的精度。
