【目标检测】算法 发展脉络 (Detection-Timeline)

Start-of-the-art Benchmark

COCO | Detection Leaderboard
这里写图片描述

Timeline

YearMonthAlgorithmImproveConference
2012AlexNetbasemodel (R-CNN、SPPNet)ImageNet2012 1st、NIPS2012
=====
201311R-CNNtraditional method -> CNN method、proposal、target methodCVPR2014
201311ZFNetbasemodel (Faster R-CNN)
=====
20146SPPNetshare feature extraction、SPP layerECCV2014
20149GoogleNetbasemodel (YOLOv1)ImageNet2014 2nd
20149VGGNetbasemodel (Fast R-CNN、Faster R-CNN、SSD)ImageNet2014 1st
=====
20152BNeasy to adjust parameters & accelerate convergence
20154Fast R-CNN4-stage -> 2-stage、RoIPooling、Softmax、Smooth L1、mini-batch=2ICCV2015
20156Faster R-CNNSS -> RPN、anchor method、new target methodNIPS2015
20156YOLOv11-stage、real-timeCVPR2016
201512SSDfeature pyramid、precise 1-stage、multi-anchor method、FCN backboneECCV2016
201512ResNetbasemodel (R-FCN、FPN、Mask R-CNN、RetinaNet)、identity mappingImageNet2015 1st、CVPR2016 Best Paper
=====
20165R-FCNshare roi-wise subnet、2-stage FCN backboneNIPS2016
20168DenseNetprecise-basemodel
201610Xceptionlight-basemodel
201611ResNeXtprecise-basemodel
201611Speed/accuracy trade-offsanalyse detection
201612YOLOv2large-scale detector、multi-scale training、joint trainingCVPR2017 Best Paper Mention
201612FPNenhanced-feature pyramidCVPR2017
=====
20171DSSDSSD+FPN
20173Mask R-CNNRoIPooling -> RoIAlignICCV2017 Best Paper
20173DCNdeformable module
20174Soft-NMSeclectic version of NMS
20174MobileNetlight-basemodel
20177ShuffleNetlight-basemodel
20177DPNbasemodelImageNet2017 1st
20178RetinaNetCE -> FL、best 1-stageICCV2017 Best Student Paper
201711MegDetlarge mini-batchCOCO2017 1st、CVPR2018
201711Light-head R-CNNlight-head
201711Relation Network
201711SNIPimage pyramidCVPR2018
201712Cascade R-CNNcascade method、2-stage -> 4-stageCVPR2018
201712R-FCN-3000precise large-scale detector
=====
20183PANFPN -> PANCOCO2017 2nd
20184YOLOv3
20184DetNetdetection-backbone
20185SNIPER
20186Understanding BNanalyse BN
  • 4
    点赞
  • 5
    收藏
    觉得还不错? 一键收藏
  • 1
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论 1
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值