Deep Learning-based Pipeline for Module Power Prediction from EL Measurements
Universität Erlangen-Nürnberg (FAU)(埃尔朗根-纽伦堡大学);Forschungszentrum Jülich GmbH(于利希研究中心有限公司);Helmholtz-Institute Erlangen-Nuremberg for Renewable Energies(亥姆霍兹埃尔朗根-纽伦堡可再生能源研究所);School of Advanced Optical Technologies, Erlangen(埃尔朗根高等光学技术学院)
Mathis Hoffmann, Claudia Buerhop-Lutz, Luca Reeb, Tobias Pickel, Thilo Winkler, Bernd Doll, Tobias Würfl, Ian Marius Peters, Christoph Brabec, Andreas Maier, Vincent Christlein
Automated inspection plays an important role in monitoring large-scale photovoltaic power plants. Commonly, electroluminescense measurements are used to identify various types of defects on solar modules but have not been used to determine the power of a module. However, knowledge of the power at maximum power point is important as well, since drops in the power of a single module can affect the performance of an entire string. By now, this is commonly determined by measurements that require to discontact or even dismount the module, rendering a regular inspection of individual modules infeasible. In this work, we bridge the gap between electroluminescense measurements and the power determination of a module. We compile a large dataset of 719 electroluminescense measurementsof modules at various stages of degradation, especially cell cracks and fractures, and the corresponding power at maximum power point. Here,we focus on inactive regions and cracks as the predominant type of defect. We set up a baseline regression model to predict the power from electroluminescense measurements with a mean absolute error of 9.0+/-3.7$W_P$ (4.0+/-8.4%). Then, we show that deep-learning can be used to train a model that performs significantly better (7.3+/-2.7$W_P$ or 3.2+/-6.5%) and propose a variant of class activation maps to obtain the per cell power loss, as predicted by the model. With this work, we aim to open a new research topic. Therefore, we publicly release the dataset, the code and trained models to empower other researchers to compare against our results. Finally, we present a thorough evaluation of certain boundary conditions like the dataset size and an automated preprocessing pipeline for on-site measurements showing multiple modules at once.
ReconfigISP: Reconfigurable Camera Image Processing Pipeline
SenseTime Research(商汤科技研究院);The Chinese University of Hong Kong(香港中文大学);Beihang University(北京航空航天大学);Nanyang Technological University(南洋理工大学);Shanghai AI Laboratory(上海人工智能实验室)
DeepShapeMatchingKit: Accelerated Functional Map Solver and Shape Matching Pipelines Revisited
Technical University of Munich(慕尼黑工业大学);Munich Center for Machine Learning(慕尼黑机器学习中心);MIT(麻省理工学院);Adobe;Imperial College London(帝国理工学院);Simon Fraser University(西蒙菲莎大学)
A Novel Deep Learning Pipeline for Retinal Vessel Detection in Fluorescein Angiography
University of Rochester(罗切斯特大学);University of Rochester Medical Center(罗切斯特大学医学中心);Houston Methodist Hospital(休斯顿卫理公会医院);Weill Cornell Medical College(威尔康奈尔医学院);Retina Consultants of Houston(休斯顿视网膜顾问中心);Blanton Eye Institute(布兰顿眼科研究所)
Segmentation of cell-level anomalies in electroluminescence images of photovoltaic modules
Vicomtech Foundation, Basque Research and Technology Alliance (BRTA)(Vicomtech 基金会,巴斯克研究与技术联盟);TECNUN School of Engineering, University of Navarra(纳瓦拉大学 TECNUN 工程学院);Institute of Data Science and Artificial Intelligence, University of Navarra(纳瓦拉大学数据科学与人工智能研究所);Tecnalia Research & Innovation, Basque Research and Technology Alliance (BRTA)(Tecnalia 研究与创新中心,巴斯克研究与技术联盟)
One Network to Solve Them All: A Sequential Multi-Task Joint Learning Network Framework for MR Imaging Pipeline
College of Computer Science, Sichuan University(四川大学计算机学院);College of Electrical Engineering, Sichuan University(四川大学电气工程学院);Chengdu University of Information Technology(成都信息工程大学)
Deep Label Fusion: A 3D End-to-End Hybrid Multi-Atlas Segmentation and Deep Learning Pipeline
University of Pennsylvania(宾夕法尼亚大学);Lund University(隆德大学)
深度学习(DL)是各种医学图像分割任务中最先进的方法。然而,它需要相对大量的手动标记的训练数据,这在某些应用中可能是不可行的。此外,DL方法对样本外数据的泛化能力相对较差。另一方面,多图谱分割(MAS)使用有限的训练数据量和良好的泛化能力具有很好的性能。一个混合的方法,集成了DL的高精度和MAS的良好的概括性是非常需要的,并可能在分割问题中发挥重要作用,手动标记的数据是很难产生的。大多数先前的工作集中在使用DL改进MAS的单个组件,而不是通过端到端管道直接优化最终的分割精度。只有一项研究探讨了这个想法在二进制分割的2D图像,但它仍然是未知的,它是否推广以及多类3D分割问题。在这项研究中,我们提出了一个3D端到端的混合管道,命名为深度标签融合(DLF),利用MAS和DL的优势。实验结果表明,DLF产生显着改善传统的标签融合方法和U-Net,直接DL方法,在使用3 T T1加权和T2加权MRI分割内侧颞叶子区域的背景下。此外,当应用于在7 T中获取的看不见的相似数据集时,DLF保持其优越的性能,这证明了其良好的推广性。
NeXtQSM -- A complete deep learning pipeline for data-consistent quantitative susceptibility mapping trained with hybrid data
The University of Queensland(昆士兰大学);Siemens Healthcare Pty Ltd(西门子医疗私人有限公司);Medical University of Vienna(维也纳医科大学);Medical University of Graz(格拉茨医科大学);Karl Landsteiner Institute for Clinical Molecular MR in Musculoskeletal Imaging(卡尔·兰德斯泰纳肌肉骨骼成像临床分子磁共振研究所);Friedrich-Alexander-Universität Erlangen-Nürnberg(弗里德里希-亚历山大-埃尔朗根-纽伦堡大学)
DeepSPV: A Deep Learning Pipeline for 3D Spleen Volume Estimation from 2D Ultrasound Images
School of Biomedical Engineering and Imaging Science, King’s College London(生物医学工程与成像科学学院,伦敦国王学院);Department of Biomedical Engineering, National University of Singapore(生物医学工程系,新加坡国立大学);School of Electronics & Computer Science, University of Southampton(电子与计算机科学学院,南安普顿大学);Ultromics Ltd(Ultromics有限公司);Evelina London Children’s Hospital, Guy’s and St Thomas NHS Foundation Trust(埃文琳伦敦儿童医院,盖茨医院及圣 Thomas NHS基金会信托)
A deep learning pipeline for breast cancer ki-67 proliferation index scoring
FEMTO-ST institute(弗朗什-孔泰电子、机械、热学与光学科学与技术研究所);Univ. Bourgogne Franche-Comte(勃艮第-弗朗什-孔泰大学);CNRS(法国国家科学研究中心);ENSMM(国立高等机械与微技术学院);Beihang University(北京航空航天大学);Hôpital Nord Franche-Comté(北弗朗什-孔泰医院);Batna2 University(巴特纳第二大学)
Ki-67增殖指数是一种重要的生物标志物,有助于病理学家诊断和选择适当的治疗方法。然而,Ki-67的自动评估是困难的,由于核重叠和复杂的变化,在其属性。本文提出了一种用于Ki-67精确自动计数的集成管道,其中突出了核分离技术的影响。首先,语义分割是通过结合挤压和激励Resnet和Unet算法从背景中提取细胞核。提取的细胞核,然后分为重叠和非重叠区域的基础上的八个几何和统计特征。随后提出了一种基于标记的分水岭算法,该算法仅适用于重叠区域以分离核。最后,使用Resnet 18从每个核块中提取深度特征,并通过随机森林分类器将其分类为阳性或阴性。在Hôpital Nord Franche-Comté医院病理科的数据集上验证了拟议管道的性能。
PrimSeq: a deep learning-based pipeline to quantitate rehabilitation training
New York University Langone Health(纽约大学朗格尼健康中心);New York University(纽约大学);Columbia University(哥伦比亚大学);Courant Institute of Mathematical Sciences, New York University(纽约大学库朗数学科学研究所)
Deep Learning for Predictive Analytics in Reversible Steganography
National Institute of Informatics(国立信息学研究所);Feng Chia University(逢甲大学);University of Jinan(济南大学);Fujian Polytechnic Normal University(福建技术师范学院);University of Warwick(华威大学);Deakin University(迪肯大学)
TinyViT: Field Deployable Transformer Pipeline for Solar Panel Surface Fault and Severity Screening
Thiagarajar College of Engineering(泰迦拉ajar工程学院);Velammal College of Engineering and Technology(韦拉马尔工程与技术学院);Kumaraguru College of Technology(库马拉加鲁技术学院)
Debiasing pipeline improves deep learning model generalization for X-ray based lung nodule detection
University of Technology Sydney(悉尼科技大学);IBM Australia Limited(IBM澳大利亚有限公司);University of New England(新英格兰大学);King Abdulaziz University(阿卜杜勒阿齐兹国王大学);Universiti Kebangsaan Malaysia(马来西亚国立大学);Charles Sturt University(查尔斯特大学);Westmead Hospital(韦斯特米德医院);Singapore University of Social Sciences(新加坡社科大学);Cogninet Australia(Cogninet澳大利亚);University of Southern Queensland(南昆士兰大学);Ngee Ann Polytechnic(义安理工学院);Asia University(亚洲大学)
SHOP: A Deep Learning Based Pipeline for near Real-Time Detection of Small Handheld Objects Present in Blurry Video
University of California San Diego(加州大学圣地亚哥分校);California State Polytechnic University, Pomona(加州州立理工大学波莫纳分校);University of Illinois at Urbana Champaign(伊利诺伊大学厄巴纳-香槟分校)
虽然先前的工作已经研究和开发了能够进行目标检测的计算模型,但是模型仍然难以可靠地解释具有运动模糊和小目标的图像。此外,这些型号都不是专门为手持式目标检测设计的。在这项工作中,我们提出了SHOP(小型手持对象管道),一个管道,可靠和有效地解释模糊的图像包含手持对象。管道每个阶段使用的特定模型是灵活的,可以根据性能要求进行更改。首先,图像被去模糊,然后通过姿势检测系统,其中在任何在场的人的手周围提出感兴趣的区域。接下来,通过单级对象检测器对图像执行目标检测。最后,所提出的感兴趣区域被用来过滤掉低置信度的检测。在Microsoft Common Objects in Context(MS COCO)的手持设备子集上进行的测试表明,这三个阶段的过程导致误报减少了70%,而在其最强配置中,误报仅减少了17%。我们还提出了一个子集的MS COCO组成的手持对象,可用于继续发展的手持目标检测方法。https://github.com/spider-sense/SHOP
FastSurfer -- A fast and accurate deep learning based neuroimaging pipeline
German Center for Neurodegenerative Diseases (DZNE)(德国神经退行性疾病中心(DZNE));A.A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital(麻省总医院A.A.马丁诺斯生物医学成像中心);Harvard Medical School(哈佛医学院);MIT(麻省理工学院)
Deep Learning Pipeline for Automated Visual Moth Monitoring: Insect Localization and Species Classification
Friedrich Schiller University Jena(耶拿弗里德里希·席勒大学);German Aerospace Center (DLR)(德国航空航天中心(DLR));Michael Stifel Center Jena for Data-Driven and Simulation Science(耶拿迈克尔·施蒂费尔数据驱动与模拟科学中心)
Fringe Projection Based Vision Pipeline for Autonomous Hard Drive Disassembly
School of Electrical and Computer Engineering, University of Georgia(佐治亚大学电气与计算机工程学院);Alcon Research Laboratories(阿康研究实验室);School of Environmental, Civil, Agricultural and Mechanical Engineering, University of Georgia(佐治亚大学环境、土木、农业和机械工程学院)