深度学习的物体识别原版论文(英文)

所需积分/C币:46 2018-12-26 22:59:44 1.33MB PDF
收藏 收藏
举报

卷积神经网络,深度学习,主动学习,半监督学习。 题目:Incremental Deep Learning for Robust Object Detection in Unknown Cluttered Environments。
Tlis article has been accepted fur publica ion in a future issue uf this journal, but has not been fully edited Content may change prior to ial publication. Citation information:DOI 10.1109/ ACCESS.2018.2875720, IEEE Access lEEEAccess multidisciplinary Rapd Review: Open Access Journal 2169-3536(C)2018 IEEE. Translat ons and content mining are permitted for academic research only. Perscnal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standardspublications/rights/index.htmlformoreinformation Tlis article has been accepted fur publica ion in a future issue uf this journal, but has not been fully edited Content may change prior to ial publication. Citation information:DOI 10.1109/ ACCESS.2018.2875720, IEEE Access lEEEAccess multidisciplinary Rapd Review: Open Access Journal dataset Tentative dataset →斗 Pseudo labeled dataset cnn based al Incremental bin based SSL selection uncertainty Batch of Detection Recognition confidence samples dataset 2169-3536(C)2018 IEEE. Translat ons and content mining are permitted for academic research only. Perscnal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standardspublications/rights/index.htmlformoreinformation Tlis article has been accepted fur publica ion in a future issue uf this journal, but has not been fully edited Content may change prior to ial publication. Citation information:DOI 10.1109/ ACCESS.2018.2875720, IEEE Access lEEEAccess multidisciplinary Rapd Review: Open Access Journal P(y, blx), Convolute Feature Convolution freeze Extractor P(,blx Detection layer Detection laver C1,C2,…,Csmn,D C1, C2, and c sunI sup C1,C2…,Cbj,Dob C, C, and c mage BB(FV) class 0…9 Sub class 2169-3536(C)2018 IEEE. Translat ons and content mining are permitted for academic research only. Perscnal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standardspublications/rights/index.htmlformoreinformation Tlis article has been accepted fur publica ion in a future issue uf this journal, but has not been fully edited Content may change prior to ial publication. Citation information:DOI 10.1109/ ACCESS.2018.2875720, IEEE Access lEEEAccess multidisciplinary Rapd Review: Open Access Journal argmin logp(! FVsup, D)+logp(FVsup, D) P(FVIL,D)=P(FV, D)p(FV ID) FVsup, FVobjFvsub +logp(llFVobj, FVsup D)+ log p(Obi Fysup d ∑[ogp(r tr Fvoi fve supr +log p(pe P(FVsup, FVobj, FVsubll, D) p(Fvsupll, Dp(Fvobilfvsup, l, dp (fvsub Fvobi, FVsup, l,D) P(FVSup D)P(FVSup, D) (FVsup, FVobill, D)=argmin f(FVsup FVob; FVsub D) + argmin f(FVobji FVsup FVsub D) P(Obi, FVsupr D)P(Vobil FVsup, D) P(IFVsub, FVob i, FVsup D)p(evsub l Obi, ISup,D F V, Fv sup (FVSum, FVobi fvsub) Algorithm 1 block Coordinate Descent-based Optimi (FVup, FVobi, FvsubMAP(L, D) Input: Image 1, d with the hierarchical tree, [p(SUp, D p(FVsup,D regularization thresholds (1, n2) argmax P(lFVobj, FVsup D)p(FVobilFYSup D) F P(Vsub, FVobi FVsup, D)P(Subl FVobj, FVsup, D )I Output:(FVsup, FVobi, fvsub Method Initialize FVsup FVobj, and FVsub part, repeat FVsup< argmin/sup(FVsup FVOhj, FVsub) Flub FVobj e argmin Jobi( FVobj; FVsup, Fvsub C FV gmin sub( FVSub FVSuD, FI (FVup, FVobj, parti., partR)MAP(L, D) P(IVU, DP(FYsup D) until convergence rmax P(IFVobj, FVsup D)p(FVobilFVsup, D) FVsupfvobjarvsub p(, FVob, FVsup D)P(part FVob Sup,D sup Job j (4) P(IFVSuD D) p(I [FVobj, FVsupr Fvsup FW al tRMAP L, D) 2169-3536(C)2018 IEEE. Translat ons and content mining are permitted for academic research only. Perscnal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standardspublications/rights/index.htmlformoreinformation Tlis article has been accepted fur publica ion in a future issue uf this journal, but has not been fully edited Content may change prior to ial publication. Citation information:DOI 10.1109/ ACCESS.2018.2875720, IEEE Access lEEEAccess multidisciplinary Rapd Review: Open Access Journal n f(r) diversity 0 Diversity={k|x∈ Tentative0≤f(X)≤1 Rank(x)< n well Rank(x) tentative well tentative data that are used to construct the CNN's pre gmax∈ diversity f(r) Xtop∈ Diversity diversity diversit op= argma∈ d diversty imaxxynxry∈D△(x1x,)} D tentative Dwell < Tentative D tentative D△< Dtentat 2169-3536(C)2018 IEEE. Translat ons and content mining are permitted for academic research only. Perscnal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standardspublications/rights/index.htmlformoreinformation Tlis article has been accepted fur publica ion in a future issue uf this journal, but has not been fully edited Content may change prior to ial publication. Citation information:DOI 10.1109/ ACCESS.2018.2875720, IEEE Access lEEEAccess f(x) Rank(r) A well Tentative D△ using f 7, X|X∈ Dtentat Rank(x=9<n Xt argmaxx c Diversity maxx,x cDd(xir x; D4|=y; U Topi D BSeq= binl BSeq(1)≠φ bini= bseq(i) u b Ac Acc(+)= maxAcc (i+1) bini acc bin= argmaxACC bin i train train u binf 1)<AcC() b f dwelL= DweLL D 2169-3536(C)2018 IEEE. Translat ons and content mining are permitted for academic research only. Perscnal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standardspublications/rights/index.htmlformoreinformation Tlis article has been accepted fur publica ion in a future issue uf this journal, but has not been fully edited Content may change prior to ial publication. Citation information:DOI 10.1109/ ACCESS.2018.2875720, IEEE Access lEEEAccess multidisciplinary Rapd Review: Open Access Journal C.5 050100150200250300350400450500 Epoch 6 0.5 30 init bin bin binbin bin binbinbin bin binbin bin binbin bin binbin bin bin bin 56789012345678 Training phase 0.2 0.15 005 1C0 00 2169-3536(C)2018 IEEE. Translat ons and content mining are permitted for academic research only. Perscnal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standardspublications/rights/index.htmlformoreinformation Tlis article has been accepted fur publica ion in a future issue uf this journal, but has not been fully edited Content may change prior to ial publication. Citation information:DOI 10.1109/ ACCESS.2018.2875720, IEEE Access lEEEAccess multidisciplinary Rapd Review: Open Access Journal Image Universe ndoc, oalAnimalPersonVehicle souce uai unning rottedsate ase Case2 0 initbin bin binbinbin bin binbin bin b n bin bin binbin bin b n binbin bin bin bin binbinbin bin bin bin binbin bi 012345678901234567890 3456739 Training Phase [.60.5,0.6][05,0.505 0 0.1 init hin bin binbinbinbin bin bin bin bin bin bin bin bin hin hin bin bin hin hin 3456789C123456789 init bin bin bin bin bin oin bin bin bin bin bin bin bin bin bin pin bin binbinbin binb in bin b n binbinbin bin bin bin 012345678901234567890123456789 Training Phase Phase 3 0.6.06 Training Phase 080.80.8]=0.6,0.8.0.8=[0.80.608-[8,0806 ■[.6.06,0.6]■D.5,0.5,0.5 0.3 目 initbin binbin bin bin bin binbin bin b n binbin binbin bin b n binbinbin bin bin bin bin binbin bin binbinbin bin 012345678951234557890123456739 Training phase [0.5,c.5,0.5]Ass=0.5,05,0.5ssL ruil bir biri biri bir biun bin tiri bir bin bin br bin bin biri bin bin birl bir bin tin 89Q 23456了89 ph≈1 Phase z Training Phase ■[08.0808]■[06.0.80.8■0.8.06,0.8■[08.0.8,0.6]■[0.6.06,0.6 2169-3536(C)2018 IEEE. Translat ons and content mining are permitted for academic research only. Perscnal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standardspublications/rights/index.htmlformoreinformation

...展开详情
试读 14P 深度学习的物体识别原版论文(英文)
立即下载 低至0.43元/次 身份认证VIP会员低至7折
    一个资源只可评论一次,评论内容不能少于5个字
    qhfei 不错的资料,可以好好学习!
    2020-07-29
    回复
    • 分享学徒

      成功上传1个资源即可获取
    关注 私信 TA的资源
    上传资源赚积分,得勋章
    最新推荐
    深度学习的物体识别原版论文(英文) 46积分/C币 立即下载
    1/14
    深度学习的物体识别原版论文(英文)第1页
    深度学习的物体识别原版论文(英文)第2页
    深度学习的物体识别原版论文(英文)第3页
    深度学习的物体识别原版论文(英文)第4页
    深度学习的物体识别原版论文(英文)第5页

    试读已结束,剩余9页未读...

    46积分/C币 立即下载 >