Computer Vision Series recent paper (attached Introduction)
Target Detection
1. Summary: Adaptation target detection depth of field Title: Deep Domain Adaptive Object Detection: a Survey Author: Wanyi Li, Peng Wang link: https: //arxiv.org/abs/2002.06797
thesis reviews the 40 Pian literature, by the Chinese Academy of Sciences Automation scholar release. Deep Learning (DL) based target detection has made great progress, these methods usually assume that a large number of tagged training data available, and the training and test data extracted from the same distribution. However, in practice these two assumptions are not always true. Deep domain adaptive target detection (DDAOD) came into being as a new learning paradigm. This paper reviews the research progress of adaptive target detection method of deep domain.
2. depth study of Abnormal detection: Summary
Title: Anomalous Instance Detection in Deep Learning: A Survey Author: Saikiran Bulusu, Dawn Song link: https: //arxiv.org/abs/2003.06979
thesis combed 119 Pian literature from Syracuse University published scholar. Discuss various examples of abnormal detection methods, and analysis of the relative advantages and disadvantages of each method.
3. The use of mobile camera detects moving objects: a comprehensive review Title: Moving Objects Detection with a Moving Camera : A Comprehensive ReviewAuthor: Marie-Neige Chapel, Thierry Bouwmans link: https: //arxiv.org/abs/2001.05238
thesis combed 347 Pian literature. With the rise of motion sensor, the camera is moved gradually become popular research direction. In this paper, various conventional methods of identification, and a plane which is divided into two or more. In both categories, the method of various types into 8 groups : panoramic background subtraction, dual cameras, motion compensation, subspace segmentation, motion segmentation, plane + parallax, multi-plane, and block-divided image. This article also publicly available data sets and evaluation index were studied.
Image Classification
4. Survey of Semi-supervised image classification, self-supervised and unsupervised
标题：A survey on Semi-, Self- and Unsupervised Techniques in Image Classification
Author: Lars Schmarje, Reinhard Koch link: https: //arxiv.org/abs/2002.08721
thesis reviews the 51 Pian literature. Summary of the less commonly used classification label image 21 is Species techniques and methods. We compare methods, and identified three major trends.
Image Denoising
5. denoising depth study: Summary Title: Deep Learning on Image Denoising: An overview of: Chunwei Tian, Chia-Wen Lin link: https: //arxiv.org/abs/1912.13171
This article reviews the 238 Pian literature by Harbin Institute of Technology, Guangdong University of Technology, Tsinghua University scholars publish. There are different types of noise processing depth learning a huge difference, but there is little research to related summary. In this paper, image noise removal at different depths learning techniques were compared, analyzed the motives and principles of different methods and compared in a common set of data de-noising. The study included: (1) CNN white noise added image; (2) for a real noise CNN image; (3) for blind noise denoising CNN; (. 4) for mixing the image noise CNN..
Image segmentation
6. Use deep learning image segmentation: Summary Title: Image Segmentation Using Deep Learning: A Survey Author: Shervin Minaee, Demetri Terzopoulos link: https: //arxiv.org/abs/2001.05566
This article reviews the 172 Pian literature, semantic examples of segmentation literature and conducted a comprehensive review, covering a variety of seminal works, including a full convolution pixel tags network encoder - decoder architecture, based on multi-scale pyramid method, recursive network, as well as against the visual attention models the generated model.
Face Recognition
DeepFakes 7. The : face detecting falsification and manipulation Summary Title: DeepFakes and Beyond: A Survey of Face Manipulation and Fake Detection of: Ruben Tolosana, Javier Ortega-Garcia
Link: https: //arxiv.org/abs/2001.00179
This paper reviews 105 Pian literature, paper facial image manipulation technologies (including DeepFake method) and a method of such detection techniques comprehensive review. It discusses the four types of facial operations: full-face synthesis, facial identity exchange (DeepFakes), facial expressions and facial attribute manipulation operations.
Pose estimation
8. goal posture review: 3D bounding box from the detector to the full 6D pose estimator Title: A Review on Object Pose Recovery: from 3D Bounding Box Detectors to Full 6D Pose Estimators Author: Caner Sahin, Tae-Kyun Kim link: https : //arxiv.org/abs/2001.10609
This article reviews the 206 Pian literature from Imperial College London published scholar. In this paper, 3D bounding box of the detector to the full 6D pose estimator objects pose recovery method conducted the first comprehensive review. Based on the mathematical model, the method is divided into various types of classification, regression, classification and regression, template matching and feature matching point for the task.
Behavior / action recognition
9. Based on the operation of the 3D skeleton identification learning method
标题：A Survey on 3D Skeleton-Based Action Recognition Using Learning Method
Author: Bin Ren, Hong Liu link: https: //arxiv.org/abs/2002.05907
This paper reviews 81 Pian literature from Beijing University published scholar. This article emphasizes the importance and necessity of 3D skeleton motion recognition data, and then data-driven manner based on recurrent neural network based on convolution neural networks and mainstream motion recognition technology based on FIG convolution network is fully described, it also the first to use 3D skeleton motion data based on a comprehensive study to identify the depth of learning.
Population count
10. Based on population density estimates and CNN's count: Summary Title: CNN-based Density Estimation and Crowd Counting: A Survey Author: Guangshuai Gao, Yunhong Wang link: https: //arxiv.org/abs/2003.12783
This paper reviews the 222 Pian literature from University of Aeronautics and Astronautics, Beijing issued scholars, CNN estimation method based on the density map, the research 220 + work on population counts were comprehensive and systematic study. At the same time according to the evaluation index, select the three best performers in the crowd statistical data sets, and analyze their strengths and weaknesses.
Medical Imaging
11. A comprehensive review title pathology image analysis using classical and deep neural networks breast tissue: A Comprehensive Review for Breast Histopathology Image Analysis Using Classical and Deep Neural Networks Author: Xiaomin Zhou, Tao Jiang link: https: // arxiv. org / abs / 2003.12255
This paper reviews the 180 Pian literature from Northeastern University published scholar. For BHIA technology based on artificial neural network is a comprehensive overview of the BHIA system is divided into classic and depth of the neural network to carry out in-depth research, analysis of existing models in order to find the most appropriate algorithm, and provide a publicly accessible data set.
12. Use deep neural network medical image registration: comprehensive overview Title: Medical Image Registration Using Deep Neural Networks : A Comprehensive Review of: Hamid Reza Boveiri, Ali Reza MehdiZadeh link: https: //arxiv.org/abs/2002.03401
article combed 117 Pian literature on the use of deep neural networks for medical image registration of the latest literature conducted a comprehensive review of the system to cover the related work in this area, including the key concepts, statistical analysis, key technology, the main contribution to the challenges and future directions.
13.Towards automated threat detection: X-ray imaging in depth safety review progress in learning Title: Towards Automatic Threat Detection: A Survey of Advances of Deep Learning within X-ray Security Imaging Author: Samet Akcay, Toby Breckon link: https: // arxiv. org / abs / 2001.01293
This paper reviews the 151 Pian literature by the British Durham University published scholar. This article is divided conventional machine learning and deep learning contemporary categories to review the safety of X-ray imaging algorithm. The depth of learning methods into supervised, semi-supervised and unsupervised learning, focuses on classification, detection, segmentation and anomaly detection task, contains a perfect X-ray data set.
14. The depth review of neural network model is used to calculate the title histopathology: Deep neural network models for computational histopathology : A survey Author: Chetan L. Srinidhi, Anne L. Martel link: https: //arxiv.org/abs/1912.12378
This article reviews the 130 Pian literature from the University of Toronto published scholar. In this paper the latest deep learning pathology image analysis method used in the organization conducted a comprehensive review, including supervision, weak supervision, unsupervised, migration and other areas of learning, and summarizes some of the existing open data sets.
Three-dimensional reconstruction
15. The external shape of the internal 3D structure prediction Summary Title: A Survey On 3D Inner Structure Prediction from its Outer Shape Author: Mohamed Mejri, Cédric Pradalier link: https: //arxiv.org/abs/2002.04571
This paper reviews 81 Pian literature from Peking University published scholar. Because the contents of the past and the little skeleton data, this paper is to study the framework for the use of 3D data in a comprehensive discussion on the motion recognition depth study of the first chapter. Article highlights the importance of identifying and 3D motion data of the skeleton, data-driven manner mainstream motion recognition technology based on recurrent neural network, the neural network convolution and convolutional networks were fully described. And describes the maximum 3D dataset skeletal NTU-RGB + D and its new version NTU-RGB + D 120, and discusses several prior top algorithms.
3D point cloud
16. No objective point cloud registration review Title: Target-less registration of point clouds : A review Author: Yue Pan
paper 48 Pian literature of the sort, summarize the basic work without destination point cloud registration, review three common types of registration methods that feature matching method based on iterative closest point algorithm and stochastic assumptions, and analyzes the advantages and disadvantages of these methods, to present their common scenario. Links: HTTPS: //arxiv.org/abs/1912.12756
OCR :
17. handwritten optical character recognition (OCR): Literature Review Integrated Systems (SLR) Title: Handwritten Optical Character Recognition (OCR) : A Comprehensive Systematic Literature Review (SLR) Author: Jamshed Memon, Rizwan Ahmed Khan link: https: // arxiv .org / abs / 2001.00139
paper 142 Pian literature carried out and summarizes the research on the OCR reviewed published between 2000 and 2018 research articles outlining the latest results and OCR technology, and analyzes the research gaps, to summarize research.
Depth depth Related:
18. Based on the depth of learning monocular depth estimation: Summary Title: Monocular Depth Estimation Based On Deep Learning : An Overview Author: Chaoqiang Zhao, Feng Qian link: https: //arxiv.org/abs/2003.06620
paper 119 Pian literature carried out and by the East China University published scholar. With the rapid development of deep neural networks, based on the depth of learning monocular depth estimation has been widely studied. In order to improve the accuracy of depth estimation, various network frameworks, loss of function and training strategies. Therefore, this paper reviews the current monocular depth estimation method based on the depth of learning, summarizes several data collection and evaluation based on the depth estimate depth study of a widely used, and reviews some of the existing representative according to different training methods methods: supervised, unsupervised and semi-supervised.
CNN
19. The overview paper convolution neural networks: analysis, application and prospect Title: A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects Author: Zewen Li, Wenjie Yang, Shouheng Peng, Fan Liu link: https: // arxiv .org / abs / 2004.02806
paper 119 Pian literature of the sort, by the East China University published scholar. This paper aims in this rapidly growing field convolutional neural network provides novel ideas and perspectives as possible, not only two-dimensional convolution, and involve one-dimensional and multi-dimensional convolution. First of all, this article briefly describes the history of CNN and CNN summarizes the development, introduction CNN classical model, which focuses on the key factors SOTA reached, and provides some rules of thumb by experimental analysis, the final one-dimensional, two-dimensional and multi-dimensional volume the product provides an overview of the application.
Visual sense / others
20. The information plane neural network classifier analysis Review Title: On Information Plane Analyses of Neural Network Classifiers - A Review Author: Bernhard C. Geiger link: HTTPS: //arxiv.org/abs/2003.09671 21. The low power consumption Overview title depth learning and computer vision methods: a Survey of methods for Low- Power deep learning and computer vision author: Abhinav Goel, George K. Thiruvathukal link: HTTPS: //arxiv.org/abs/2003.11066
22. the depth learning encounter when data alignment: the depth of your home network (DRN) Review title: when deep Learning Meets data alignment: A Review on deep registration networks (DRNs) author: Victor Villena-Martinez, Robert B. Fisher link: https: //arxiv.org /abs/2003.03167
23. the unlimited palmprint recognition for consumer devices: a literature review title: Towards Unconstrained palmprint recognition on consumer devices : a literature review author:. Adrian-S Ungureanu, Peter Corcoran link: https: //arxiv.org /abs/2003.00737
24. The ground-based texture localization function - Summary Title: Features for Ground Texture Based Localization - A Survey Author: Jan Fabian Schmid, Rudolf Mester link: https: //arxiv.org/abs/2002.11948
25. From the watch movement: visual indoor navigation (VIN) a Summary Title: From Seeing to Moving: A Survey on Learning for Visual Indoor Navigation (VIN) Author: Xin Ye, Yezhou Yang link: https: //arxiv.org/ abs / 2002.11310