45篇，CVPR 2019 Best paper Finalist！最佳论文入围合集（含论文地址与代码链接）

2019 年 8 月 11 日 极市平台

加入极市专业CV交流群，与6000+来自腾讯，华为，百度，北大，清华，中科院等名企名校视觉开发者互动交流！更有机会与李开复老师等大牛群内互动！

同时提供每月大咖直播分享、真实项目需求对接、干货资讯汇总，行业技术交流。点击文末“阅读原文”立刻申请入群~

三大世界顶级计算机视觉会议之一「计算机视觉与模式识别会议」（Conference on Computer Vision and Pattern Recognition 2019，CVPR 2019）在美国长滩落下帷幕，顶会吸引全球超过9200位顶尖专家、学者以及产业界人士，共同推进 CV 技术的发展与落地。

根据 CVPR2019官方网站公布的论文的最终录用结果，极市统计汇总了那些最佳论文奖项的入围者，以及同时附上论文的链接与代码地址（如果有）。虽然没有获奖，但入围论文依然可以称得上是本次 CVPR会议上投稿论文中的精华部分，一起来了解一下吧 ~

1 “Probabilistic Permutation Synchronization using the RiemannianStructure of the Birkhoff Polytope,”

利用Birkhoff多面体的黎曼结构进行概率置换同步

作者：Tolga Birdal; UmutSimsekli

论文地址：https://arxiv.org/abs/1904.05814

2 “DeepC𝑂^3: Deep Instance Co-segmentation by Co-peak Search and Co-saliencyDetection,”

DeepCO 3：通过共峰搜索和共显性检测进行深度实例共同分割

作者：Kuang-JuiHsu; Yen-Yu Lin; Yung-Yu Chuang

代码链接：https://github.com/KuangJuiHsu/DeepCO3?

3 “SelFlow: Self-Supervised Learning ofOptical Flow,”

SelFlow：光流的自我监督学习

作者：PengpengLiu, Michael Lyu, Irwin King, Jia Xu

代码链接：https://github.com/ppliuboy/SelFlow

论文地址：https://arxiv.org/abs/1904.09117

4 “SDRSAC: Semidefinite-Based RandomizedApproach for Robust Point Cloud Registration Without Correspondences,”

SDRSAC：基于半定的随机方法，用于无通信的鲁棒点云注册

作者：HuuLe; Thanh-Toan Do; Tuan NA Hoang; Ngai-Man Cheung

代码链接：https://github.com/intellhave/SDRSAC

论文地址：https://arxiv.org/abs/1904.03483

5 “Deep Tree Learning for Zero-Shot FaceAnti-Spoofing,”

用于零镜头防欺骗的深度学习

作者：YaojieLiu; Joel Stehouwer; Amin Jourabloo; Xiaoming Liu

论文地址：https://arxiv.org/abs/1904.02860

6 “Neural RGB->D Sensing: Depth andUncertainty from a Video Camera,”

神经 RGB - > d 传感：深度和不确定性从一个摄像机

作者：Chao Liu; Jinwei Gu; KihwanKim; Srinivasa G Narasimhan; Jan Kautz

论文地址：https://arxiv.org/abs/1901.02571

7 “Self-Supervised 3D Hand Pose Estimation ThroughTraining by Fitting ,”

基于拟合训练的自监督三维手势估计

作者：Chengde Wan; Thomas Probst; Luc Van Gool; Angela Yao

代码链接：https://github.com/melonwan/sphereHand

8 “Neural Illumination: Lighting Predictionfor Indoor Environments,”

神经照明：室内环境的照明预测

作者：ShuranSong; Thomas Funkhouser

论文地址：https://arxiv.org/abs/1906.07370

9 “Shapes and Context: In-the-wild ImageSynthesis & Manipulation,”

形状和背景：野外图像合成和操作

作者：AayushBansal; Yaser Sheikh; Deva Ramanan

论文地址：https://arxiv.org/abs/1906.04728

10 “SiCloPe: Silhouette-based Clothed People,”

SiCloPe：基于剪影的穿着衣服的人

作者：RyotaNatsume; ShunsukeSaito; Zeng Huang; WeikaiChen; ChongyangMa;Shigeo Morishima; Hao Li

论文地址：https://arxiv.org/abs/1901.00049

11 “A General and Adaptive Robust LossFunction,”

一般自适应鲁棒损失函数

作者：Jonathan T Barron

代码地址：https://github.com/jonbarron/robust_loss_pytorch

论文地址：https://arxiv.org/abs/1701.03077

12 “2.5D Visual Sound,”

2.5D视觉声音

作者：RuohanGao; Kristen Grauman

代码链接：https://github.com/facebookresearch/FAIR-Play

论文链接：https://arxiv.org/abs/1812.04204

13 “Incremental Object Learning FromContiguous Views,”

从相邻视图进行增量对象学习

作者：Stefan Stojanov; Samarth Mishra; Ngoc Anh Thai; Nikhil Dhanda; AhmadHumayun; Linda Smith; Chen Yu; James Rehg

14 “Text2Scene: Generating Compositional Scenes from TextualDescriptions,”

Text2Scene：从文本描述生成合成场景

作者：FuwenTan; Song Feng; Vicente Ordonez

代码链接：https://github.com/uvavision/Text2Image%EF%BC%9F

论文地址：https://arxiv.org/abs/1809.01110

15 “Relational Action Forecasting,”

关系行动预测

作者：Chen Sun; Abhinav Shrivastava; Carl Vondrick; Rahul Sukthankar;Kevin Murphy; Cordelia Schmid

论文地址：https://arxiv.org/abs/1904.04231

16 “Shifting More Attention to Video SalientObject Detection,”

更加注重视频显着对象检测

作者：Deng-Ping Fan;WenguanWang; Ming-Ming Cheng; JianbingShen

代码链接：https://github.com/DengPingFan/DAVSOD

17 “GA-Net:Guided Aggregation Net for End-To-End Stereo Matching,”

GA-Net：用于端到端立体匹配的引导聚合网络

作者：FeihuZhang; Victor Prisacariu; Yang Ruigang; Philip Torr

代码链接： https://github.com/feihuzhang/GANet

论文地址：https://arxiv.org/abs/1904.06587

18 “ASkeleton-bridged Deep Learning Approach for Generating Meshes of ComplexTopologies from Single RGB Images,”

一种用于从单个RGB图像生成复杂拓扑网格的骨架桥接深度学习方法

作者：JiapengTang; Xiaoguang Han; JunyiPan; KuiJia; Xin Tong

论文地址：https://arxiv.org/abs/1903.04704

19 “Semantic Image Synthesis withSpatially-Adaptive Normalization,”

空间自适应归一化的语义图像合成

作者：TaesungPark; Ming-Yu Liu; Ting-Chun Wang; Jun-Yan Zhu

链接地址：https://github.com/NVlabs/SPADE

论文地址：https://arxiv.org/abs/1903.07291

20 “ContactDB: Analyzing and Predicting GraspContact via Thermal Imaging,”

ContactDB：通过热成像分析和预测抓取接触

作者：Samarth Brahmbhatt; CusuhHam; Charlie Kemp; James Hays

代码链接：https://github.com/samarth-robo/contactdb_prediction

论文地址：https://arxiv.org/abs/1904.06830

21 “RevealingScenes by Inverting Structure from Motion Reconstructions,”

通过从运动重建中反转结构来揭示场景

作者：Francesco Pittaluga; Sanjeev J Koppal; Sing Bing Kang; Sudipta Sinha

论文地址：https://arxiv.org/abs/1904.03303

22 “A Theory of Fermat Paths for Non-Line-Of-SightShape Reconstruction,”

非视线形状重建的费马路径理论

作者：ShumianXin; Sotiris Nousias; Kyros Kutulakos; AswinSankaranarayanan;Srinivasa G Narasimhan; IoannisGkioulekas

论文地址：https://www.ri.cmu.edu/publications/a-theory-of-fermat-paths-for-non-line-of-sight-shape-reconstruction/

23 “Relation-Shape Convolutional NeuralNetwork for Point Cloud Analysis,”

用于点云分析的关系形状卷积神经网络

作者：YongchengLiu; Bin Fan; ShimingXiang; ChunhongPan

代码链接：https://github.com/Yochengliu/Relation-Shape-CNN

论文地址：https://arxiv.org/abs/1904.07601

微信解读：https://mp.weixin.qq.com/s/KuoHFPpUYHPFacsBWZWVSg

24 “BubbleNets:Learning to Select the Guidance Frame in Video Object Segmentation by DeepSorting Frames,”

BubbleNets：学习通过深度排序帧在视频对象分割中选择指导框架

作者：Brent Griffin; Jason J Corso

代码链接：https://github.com/griffbr/BubbleNets

论文地址：https://arxiv.org/abs/1903.11779

25 “ImageDeformation Meta-Networks for One-Shot Learning,”

用于一次性学习的图像变形元网络

作者：ZitianChen; Yanwei Fu; Yu-Xiong Wang; Lin Ma; Wei Liu; MartialHebert

论文地址：https://arxiv.org/abs/1905.11641

26 “Estimating 3D Motion and Forces of Person-ObjectInteractions from Monocular Video,”

从单目视频估计人体 - 物体相互作用的三维运动和力

作者：ZongmianLi; Jiri Sedlar; Justin Carpentier; Ivan Laptev; NicolasMansard; Josef Sivic

论文地址：https://arxiv.org/abs/1904.02683

27“A Style-Based Generator Architecture forGenerative Adversarial Networks,”

基于样式的生成对抗网络生成器体系结构

作者：Tero Karras; SamuliLaine; Timo Aila

代码链接：https://github.com/NVlabs/stylegan

论文地址：https://arxiv.org/abs/1812.04948

28 “Unsupervised Part-Based Disentangling ofObject Shape and Appearance,”

基于无监督部分的物体形状和外观解缠结

作者：Dominik Lorenz; Leonard Bereska; Timo Milbich; Bjorn Ommer

论文地址：https://arxiv.org/abs/1903.06946

29 “Pushing the Boundaries of ViewExtrapolation with Multiplane Images,”

用多平面图像推动视图外推边界

作者：Pratul Srinivasan; Richard Tucker; Jonathan T Barron; RaviRamamoorthi; Ren Ng; Noah Snavely

论文地址：https://arxiv.org/abs/1905.00413

30 “Path-Invariant Map Networks,”

路径不变映射网络

作者：ZaiweiZhang; ZhenxiaoLiang; LemengWu; Xiaowei Zhou; Qixing Huang

代码链接：https://github.com/zaiweizhang/path_invariance_map_network

论文地址：https://arxiv.org/abs/1812.11647

31 “Learning the Depths of Moving People byWatching Frozen People,”

通过观察冰冻人物来学习移动人的深度

作者：ZhengqiLi; Tali Dekel; Forrester Cole; Richard Tucker; Ce Liu; BillFreeman; Noah Snavely

论文地址：https://arxiv.org/abs/1904.11111

32 “Efficient Online Multi-Person 2D PoseTracking with Recurrent Spatio-Temporal Affinity Fields,”

具有循环时空亲和场的高效在线多人2D姿态跟踪

作者：YaadhavRaaj; Haroon Idrees; GinesHidalgo Martinez; Yaser Sheikh

论文地址：https://arxiv.org/abs/1811.11975

33 “Learning to Compose Dynamic TreeStructures for Visual Contexts,”

学习为视觉上下文构建动态树结构

作者：KaihuaTang; HanwangZhang; BaoyuanWu; WenhanLuo; Wei Liu

代码链接：https://github.com/KaihuaTang/VCTree-Visual-Question-Answering

论文地址：https://arxiv.org/abs/1812.01880

34 “Cascaded Projection: End-to-End NetworkCompression and Acceleration,”

级联投影：端到端网络压缩和加速

作者：Breton L Minnehan; Andreas Savakis

论文地址：https://arxiv.org/abs/1903.04988

35“Taking a Deeper Look at the InverseCompositional Algorithm,”

深入研究逆组合算法

作者：ZhaoyangLv; Frank Dellaert; James Rehg; Andreas Geiger

代码链接：https://github.com/lvzhaoyang/DeeperInverseCompositionalAlgorithm

论文地址：https://arxiv.org/abs/1812.06861

36 “Occupancy Networks: Learning 3D Reconstruction in Function Space,”

占用网络：学习功能空间中的三维重建

作者：Lars M Mescheder; Michael Oechsle; Michael Niemeyer; SebastianNowozin (Google AI Berlin); Andreas Geiger

代码链接：https://github.com/autonomousvision/occupancy_networks

论文地址：https://arxiv.org/abs/1812.03828

37 “Geometry-Consistent Generative AdversarialNetworks for One-Sided Unsupervised Domain Mapping,”

单边无监督域映射的几何一致生成对抗网络

作者：Huan Fu; MingmingGong; ChaohuiWang; Kayhan Batmanghelich; Kun Zhang;DachengTao

论文地址：https://arxiv.org/abs/1809.05852

38 “Convolutional Mesh Regression forSingle-Image Human Shape Reconstruction,”

单像人体形状重建的卷积网格回归

作者：Nikos Kolotouros; Georgios Pavlakos; Kostas Daniilidis

代码链接：https://github.com/nkolot/GraphCMR

论文地址：https://arxiv.org/abs/1905.03244

39 “Neural Rerendering in the Wild,”

野外神经再造

作者：Moustafa Meshry; Ricardo Martin-Brualla; Noah Snavely; Hugues Hoppe;SamehKhamis; Rohit Pandey; Dan B Goldman

代码链接：https://github.com/google/neural_rerendering_in_the_wild

论文地址：https://arxiv.org/abs/1904.04290

40 “Content Authentication for Neural ImagingPipelines: End-to-end Optimization of Photo Provenance in Complex DistributionChannels,”

神经成像管道的内容认证：复杂分销渠道中照片来源的端到端优化

作者：Pawel Korus; Nasir Memon

代码链接：https://github.com/pkorus/neural-imaging

论文地址：https://arxiv.org/abs/1812.01516

41 “Reinforced Cross-Modal Matching andSelf-Supervised Imitation Learning for Vision-Language Navigation,”

视觉语言导航的强化跨模态匹配和自监督模拟学习

作者：Xin Wang; QiuyuanHuang; AsliCelikyilmaz; JianfengGao; DinghanShen;Yuan-Fang Wang; William Yang Wang; Lei Zhang

论文地址：https://arxiv.org/abs/1811.10092

微信解读：https://mp.weixin.qq.com/s/LsHWkdwqqrOPFgCNNcBdpg

42 “FilterReg: Robust and EfficientProbabilistic Point-Set Registration using Gaussian Filter and TwistParameterization,”

FilterReg：使用高斯滤波器和扭曲参数化的鲁棒且高效的概率点集注册

作者：Wei Gao; Russ Tedrake

代码链接：https://bitbucket.org/gaowei19951004/poser/src/master/

论文地址：https://arxiv.org/abs/1811.10136

43 “Locating Objects Without Bounding Boxes,”

定位没有边框的对象

作者：Javier Ribera; David Güera; Yuhao Chen; Edward Delp

代码链接：https://github.com/javiribera/locating-objects-without-bboxes

论文地址：https://arxiv.org/abs/1806.07564

44 “DeepSDF: Learning Continuous SignedDistance Functions for Shape Representation,”

DeepSDF：学习形状表示的连续符号距离函数

作者：Jeong Joon Park; Peter R Florence; Julian Straub; Richard Newcombe;Steven Lovegrove

代码链接：https://github.com/Oktosha/DeepSDF-explained?

论文地址：https://arxiv.org/abs/1901.05103

45 “CollaGAN: Collaborative GAN for MissingImage Data Imputation,”

CollaGAN：缺少图像数据插补的协作GAN

作者：DongwookLee; Junyoung Kim; Won-Jin Moon; Jong Chul Ye

论文地址：https://arxiv.org/abs/1901.09764

最后附上CVPR2019/CVPR2018/CVPR2017相关论文，代码，解读资料开源项目: https://github.com/extreme-assistant/cvpr2019, 欢迎Star~

-完-

*延伸阅读

添加极市小助手微信（ID : cv-mart），备注：研究方向-姓名-学校/公司-城市（如：目标检测-小极-北大-深圳），即可申请加入目标检测、目标跟踪、人脸、工业检测、医学影像、三维&SLAM、图像分割等极市技术交流群，更有每月大咖直播分享、真实项目需求对接、干货资讯汇总，行业技术交流，一起来让思想之光照的更远吧~