本文盘点ECCV 2020 与目标检测相关的研究,包含目标检测新范式、密集目标检测、点云目标检测、少样本目标检测、水下目标检测、域适应目标检测、弱监督目标检测、训练策略等,总计 41 篇,其中 2 篇 Oral,6 篇 Spotlight,开源或者将开源的有26篇。
其中有众多非常值得参考的工作,比如Facebook的DETR,还有两篇基于训练策略无痛涨点的方法。
下载包含这些论文的 ECCV 2020 所有论文:
目标检测新范式
End-to-End Object Detection with Transformers作者 | Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko论文 | https://arxiv.org/abs/2005.12872
代码 | https://github.com/facebookresearch/detr (目前已有4.8K星)GeoGraph: Graph-based multi-view object detection with geometric cues end-to-end作者 | Ahmed Samy Nassar, Stefano D'Aronco, Sébastien Lefèvre, Jan D. Wegner
单位 | IRISA, Universite Bretagne Sud;苏黎世联邦理工学院论文 | https://arxiv.org/abs/2003.10151基于图的检测方法在城市多视角目标检测中的应用,在精度和效率方面都好于之前的方法。
UFO²: A Unified Framework towards Omni-supervised Object Detection作者 | Zhongzheng Ren , Zhiding Yu, Xiaodong Yang , Ming-Yu Liu,Alexander G. Schwing, Jan Kautz论文 | https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123640290.pdf这篇论文从最大限度利用数据集的标注出发,提出一种既能利用目标检测包围框标注,又能利用弱监督标注的目标检测训练统一框架,在实际应用中能最大限度利用训练样本,感觉这是个很好的想法。密集目标检测
BorderDet: Border Feature for Dense Object Detection作者 | Han Qiu, Yuchen Ma, Zeming Li, Songtao Liu, Jian Sun
论文 | https://arxiv.org/abs/2007.11056
代码 | https://github.com/Megvii-BaseDetection/BorderDet该文提出了一种非常简单、高效的操作来提取物体边界极限点的特征,叫做“BorderAlign”。模型只增加很少的时间开销,可以在经典模型上实现FCOS(38.6 v.s. 41.4). FPN(37.1 v.s. 40.7)。Anchor-free 目标检测
Corner Proposal Network for Anchor-free, Two-stage Object Detection作者 | Kaiwen Duan, Lingxi Xie, Honggang Qi, Song Bai, Qingming Huang, Qi Tian
论文 | https://arxiv.org/abs/2007.13816
代码 | https://github.com/Duankaiwen/CPNDet(即将)目标检测错误分析工具
TIDE: A General Toolbox for Identifying Object Detection Errors作者 | Daniel Bolya, Sean Foley, James Hays, Judy Hoffman
论文 | https://arxiv.org/abs/2008.08115代码 | https://github.com/dbolya/tide主页 | https://dbolya.github.io/tide/多目标检测和跟踪
Chained-Tracker: Chaining Paired Attentive Regression Results for End-to-End Joint Multiple-Object Detection and Tracking作者 | Jinlong Peng, Changan Wang, Fangbin Wan, Yang Wu, Yabiao Wang, Ying Tai, Chengjie Wang, Jilin Li, Feiyue Huang, Yanwei Fu
论文 | https://arxiv.org/abs/2007.14557
代码 | https://github.com/pjl1995/CTracker带方向目标的检测
PIoU Loss: Towards Accurate Oriented Object Detection in Complex Environments作者 | Zhiming Chen, Kean Chen, Weiyao Lin, John See, Hui Yu, Yan Ke, Cong Yang
单位 | 扩博智能Clobotics;上海交通大学;多媒体大学论文 | https://arxiv.org/abs/2007.09584代码 | https://github.com/clobotics/piouArbitrary-Oriented Object Detection with Circular Smooth Label作者 | Xue Yang, Junchi Yan
论文 | https://arxiv.org/abs/2003.05597
代码 | https://github.com/Thinklab-SJTU/CSL_RetinaNet目标检测定位提精
Side-Aware Boundary Localization for More Precise Object Detection作者 | Jiaqi Wang, Wenwei Zhang, Yuhang Cao, Kai Chen, Jiangmiao Pang, Tao Gong, Jianping Shi, Chen Change Loy, Dahua Lin
单位 | 香港中文大学;南洋理工大学;商汤;浙大;国科大论文 | https://arxiv.org/abs/1912.04260
代码 | https://github.com/open-mmlab/mmdetection弱监督目标检测
Many-shot from Low-shot: Learning to Annotate using Mixed Supervision for Object Detection作者 | Carlo Biffi, Steven McDonagh, Philip Torr, Ales Leonardis, Sarah Parisot
单位 | 华为;Mila Montr´eal;牛津大学论文 | https://arxiv.org/abs/2008.09694该文提出一种在线样本标注方法,可在任意目标检测算法训练中利用数据弱监督信息扩充训练样本,在Faster RCNN的实验中,分别取得了17% mAP, 9% AP50 提升在PASCAL VOC 2007 、 MS-COCO数据集上。Enabling Deep Residual Networks for Weakly Supervised Object Detection作者 | Yunhang Shen, Rongrong Ji , Yan Wang, Zhiwei Chen, Feng Zheng ,Feiyue Huang , Yunsheng Wu
单位 | 厦门大学;Pinterest;南科大;腾讯优图(上海)论文 | https://www.ecva.net/papers/eccv_2020/papers_ECCV/代码 | https://github.com/shenyunhang/DRN-WSODBoosting Weakly Supervised Object Detection with Progressive Knowledge Transfer论文 | https://arxiv.org/abs/2007.07986代码 | https://github.com/mikuhatsune/wsod_transferCheaper Pre-training Lunch: An Efficient Paradigm for Object Detection作者 | Dongzhan Zhou, Xinchi Zhou, Hongwen Zhang, Shuai Yi, Wanli Ouyang
单位 | 悉尼大学,商汤CV研究小组;中科院&国科大;商汤论文 | https://arxiv.org/abs/2004.12178
该文发明了一种计算代价小却能改进目标检测精度的预训练方法Montage pre-training。目标检测、实例分割、姿态估计全家桶
Point-Set Anchors for Object Detection, Instance Segmentation and Pose Estimation作者 | Fangyun Wei, Xiao Sun, Hongyang Li, Jingdong Wang, Stephen Lin
论文 | https://arxiv.org/abs/2007.02846
代码 | https://github.com/FangyunWei/PointSetAnchor点云目标检测
SPOT: Selective Point Cloud Voting for Better Proposal in Point Cloud Object Detection作者 | Hongyuan Du, Linjun Li, Bo Liu, and Nuno Vasconcelos
论文 | https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123560222.pdf
Streaming Object Detection for 3-D Point Clouds作者 | Wei Han, Zhengdong Zhang, Benjamin Caine, Brandon Yang, Christoph Sprunk, Ouais Alsharif, Jiquan Ngiam, Vijay Vasudevan, Jonathon Shlens, Zhifeng Chen论文 | https://arxiv.org/abs/2005.01864Pillar-based Object Detection for Autonomous Driving作者 | Yue Wang, Alireza Fathi, Abhijit Kundu, David Ross, Caroline Pantofaru, Thomas Funkhouser, Justin Solomon论文 | https://arxiv.org/abs/2007.10323代码 | https://github.com/WangYueFt/pillar-odSSN: Shape Signature Networks for Multi-class Object Detection from Point Clouds作者 | Xinge Zhu Yuexin Ma Tai Wang Yan Xu Jianping Shi Dahua Lin论文 | https://arxiv.org/abs/2004.02774代码 | https://github.com/xinge008/SSN(尚未开源)一阶段 vs. 二阶段
MimicDet: Bridging the Gap Between One-Stage and Two-Stage Object Detection作者 | Xin Lu, Quanquan Li , Buyu Li , and Junjie Yan
论文 | https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123590528.pdf通过使一阶段的目标检测方法“模仿”二阶段目标检测算法的特征,提升一阶段目标检测的精度。
无痛涨点之训练策略改进
Dynamic R-CNN: Towards High Quality Object Detection via Dynamic Training作者 | Hongkai Zhang, Hong Chang, Bingpeng Ma, Naiyan Wang, Xilin Chen
论文 | https://arxiv.org/abs/2004.06002
代码 | https://github.com/hkzhang95/DynamicRCNN该文提出Dynamic R-CNN根据训练期间proposals的统计信息自动调整IoU阈值和回归损失函数的shape(SmoothL1 Loss的参数),更好地利用了训练样本,并推动了检测器去适应更多高质量的样本。使用ResNet-50-FPN方法,没有任何其他开销,在MS COCO数据集上实现了1.9%的AP和5.5%的AP90改进。
LabelEnc: A New Intermediate Supervision Method for Object Detection作者 | Miao Hao, Yitao Liu, Xiangyu Zhang, Jian Sun论文 | https://arxiv.org/abs/2007.03282代码 | https://github.com/megvii-model/LabelEnc少样本目标检测
Multi-Scale Positive Sample Refinement for Few-Shot Object Detection作者 | Jiaxi Wu, Songtao Liu, Di Huang, Yunhong Wang
论文 | https://arxiv.org/abs/2007.09384
代码 | https://github.com/jiaxi-wu/MPSRFew-Shot Object Detection and Viewpoint Estimation for Objects in the Wild作者 | Yang Xiao, Renaud Marlet
单位 | Univ Gustave Eiffel;valeo.ai论文 | https://arxiv.org/abs/2007.12107
代码 | https://github.com/YoungXIAO13/FewShotDetection主页 | http://imagine.enpc.fr/~xiaoy/FSDetView/水下目标检测
Dual Refinement Underwater Object Detection Network作者 | Baojie Fan, Wei Chen, Yang Cong, Jiandong Tian论文 | https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123650273.pdf代码 | https://github.com/Peterchen111/FERNet(即将)目标检测数据集
APRICOT: A Dataset of Physical Adversarial Attacks on Object Detection作者 | A. Braunegg, Amartya Chakraborty, Michael Krumdick, Nicole Lape, Sara Leary, Keith Manville, Elizabeth Merkhofer, Laura Strickhart, Matthew Walmer论文 | https://arxiv.org/abs/1912.08166Object Detection with a Unified Label Space from Multiple Datasets作者 | Xiangyun Zhao, Samuel Schulter, Gaurav Sharma, Yi-Hsuan Tsai, Manmohan Chandraker, Ying Wu
单位 | 西北大学;NEC Labs America;加利福尼亚大学圣迭戈分校论文 | https://arxiv.org/abs/2008.06614
数据集 | http://www.nec-labs.com/~mas/UniDet/resources/主页 | http://www.nec-labs.com/~mas/UniDet/数据集如此珍贵,如何把不同标注的数据集结合起来使用?该文提出一种统一目标检测的标注空间。
快速训练方法
Large Batch Optimization for Object Detection: Training COCO in 12 Minutes作者 | Tong Wang , Yousong Zhu , Chaoyang Zhao , Wei Zeng , Yaowei Wang,Jinqiao Wang,Ming Tang单位 | 中科院;国科大;ObjectEye(视语科技);北大;鹏城实验室等论文 | https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123660477.pdf解读 | 12分钟训练COCO模型!速度精度双提高,https://mp.weixin.qq.com/s/pYxVPZk3H2gZQSzxKQOOQg
自监督学习+目标检测
Improving Object Detection with Selective Self-supervised Self-training作者 | Yandong Li, Di Huang, Danfeng Qin, Liqiang Wang, Boqing Gong论文 | https://arxiv.org/abs/2007.09162数据增广
Learning Data Augmentation Strategies for Object Detection作者 | Barret Zoph, Ekin D. Cubuk, Golnaz Ghiasi, Tsung-Yi Lin, Jonathon Shlens, Quoc V. Le论文 | https://arxiv.org/abs/1906.11172代码 | https://github.com/tensorflow/tpu/tree/master/models/official/detection域适应目标检测
Domain Adaptive Object Detection via Asymmetric Tri-way Faster-RCNN作者 | Zhenwei He, Lei Zhang单位 | Learning Intelligence & Vision Essential (LiVE) Group,重庆大学论文 | https://arxiv.org/abs/2007.01571代码 | https://github.com/He-Zhenwei/ATFPrior-based Domain Adaptive Object Detection for Hazy and Rainy Conditions作者 | Vishwanath A. Sindagi, Poojan Oza, Rajeev Yasarla, Vishal M. Patel
论文 | https://arxiv.org/abs/1912.00070
Collaborative Training between Region Proposal Localization and Classification for Domain Adaptive Object Detection作者 | Ganlong Zhao, Guanbin Li , Ruijia Xu, and Liang Lin
论文 | https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123630086.pdf
代码 | https://github.com/GanlongZhao/CST_DA_detection(即将)改进的新思路
Dive Deeper Into Box for Object Detection作者 | Ran Chen, Yong Liu, Mengdan Zhang, Shu Liu, Bei Yu, Yu-Wing Tai单位 | 香港中文大学;腾讯优图实验室;视摩智慧SmartMore;香港科技大学论文 | https://arxiv.org/abs/2007.14350Hierarchical Context Embedding for Region-based Object Detection作者 | Zhao-Min Chen, Xin Jin, Borui Zhao, Xiu-Shen Wei, Yanwen Guo论文 | https://arxiv.org/abs/2008.01338AABO: Adaptive Anchor Box Optimization for Object Detection via Bayesian Sub-sampling作者 | Wenshuo Ma, Tingzhong Tian, Hang Xu, Yimin Huang, Zhenguo Li
论文 | https://arxiv.org/abs/2007.09336
代码 | https://github.com/wwdkl/AABOSoft Anchor-Point Object Detection作者 | Chenchen Zhu, Fangyi Chen, Zhiqiang Shen, Marios Savvides
论文 | https://arxiv.org/abs/1911.12448
代码 | https://github.com/xuannianz/SAPDOS2D: One-Stage One-Shot Object Detection by Matching Anchor Features作者 | Anton Osokin, Denis Sumin, Vasily Lomakin
单位 | 俄罗斯国家研究型高等经济大学;Yandex;mirum.io论文 | https://arxiv.org/abs/2003.06800
代码 | https://github.com/aosokin/os2dQuantum-soft QUBO Suppression for Accurate Object Detection作者 | Junde Li, Swaroop Ghosh论文 | https://arxiv.org/abs/2007.13992Adaptive Object Detection with Dual Multi-Label Prediction作者 | Zhen Zhao, Yuhong Guo, Haifeng Shen, Jieping Ye论文 | https://arxiv.org/abs/2003.12943
HoughNet: Integrating near and long-range evidence for bottom-up object detection作者 | Nermin Samet, Samet Hicsonmez, Emre Akbas单位 | 中东技术大学;Hacettepe University论文 | https://arxiv.org/abs/2007.02355代码 | https://github.com/nerminsamet/houghnetProbabilistic Anchor Assignment with IoU Prediction for Object Detection作者 | Kang Kim, Hee Seok Lee单位 | XL8 Inc;Qualcomm Korea YH论文 | https://arxiv.org/abs/2007.08103代码 | https://github.com/kkhoot/PAA
推荐阅读
ACCV 2020国际细粒度网络图像识别竞赛即将开赛!添加极市小助手微信(ID : cvmart2),备注:姓名-学校/公司-目标检测-城市(如:小极-北大-目标检测-深圳),即可申请加入极市目标检测等技术交流群:每月大咖直播分享、真实项目需求对接、求职内推、算法竞赛、干货资讯汇总、与 10000+来自港科大、北大、清华、中科院、CMU、腾讯、百度等名校名企视觉开发者互动交流~
觉得有用麻烦给个在看啦~