Learning Dense 3D Correspondence
TPO16词汇

TPO16-1 Trade and the Ancient Middle East词mainstay 音标[ˈmenˌste]中文解析n. 支柱,柱石;骨干,中流砥柱;主要的依靠;主桅支索词caravan音标[ˈkærəˌvæn]中文解析n. (可供居住的)拖车,大篷车;(穿过沙漠地带的)旅行队(如商队);<英>活动住宅;<美>移民列车vi. 乘拖车度假;参加旅行队旅行词negotiate音标[nɪˈɡoʃiˌet]中文解析vi. 谈判,协商,交涉vt. 谈判达成;成功越过;议价出售词geology音标[dʒi'ɔlədʒi]中文解析n. 地质学;地质情况词limestone音标['laimstəun]中文解析n. [岩] 石灰岩词deposit音标[dɪ'pɑzɪt]中文解析vt. 1. 放下,放置;寄存[O] 2. 使沉淀;使沉积 3. 把(钱)储存,存放(银行等) vi. 1. 沉淀n. 1. 存款 2. 保证金;押金;定金[S1] 3. 沉淀物;矿床 4. 堆积;沉淀词metallic音标[mi'tælik, me-]中文解析adj. 金属的,含金属的词obsidian音标[ɑbˈsɪdiən]中文解析n. 黑曜石词gem音标[dʒɛm]中文解析n. 宝石,珍宝;精华;受人重视者vi. 点缀;用宝石装饰;饰以宝石adj.最佳品质的词artisan音标[,ɑ:ti'zæn, 'ɑ:tizən]中文解析n. 工匠,技工词tutelage 音标['tju:tilidʒ]中文解析n. 监护;指导词blur音标[blə:]中文解析vt. 涂污;使…模糊不清;使暗淡;玷污vi. 沾上污迹;变模糊n. 污迹;模糊不清的事物词guild音标[ɡild]中文解析n. 协会,行会;同业公会词mutual 音标['mju:tʃuəl, -tjuəl]中文解析adj. 共同的;相互的,彼此的词repudiate音标[rɪˈpjudiˌeɪt]中文解析vt. (正式地)否认;拒绝接受;拒绝与……往来;拒不履行(法律义务)词conceptualize音标[kən'septjuəlaiz]中文解析vt. 使概念化vi. 概念化词craft音标[krɑ:ft, kræft]中文解析n. 工艺;手艺;太空船vt. 精巧地制作词egalitarian音标[ɪˌɡælɪˈtɛriən]中文解析adj. 主张平等的;平等主义的n. 平等主义;平等主义者词maintenance音标[ˈmentənəns]中文解析n. 维持,保持;保养,保管;维护;维修词confederacy音标[kən'fɛdərəsɪ]中文解析n. 联盟;联邦;私党词multiplicity 音标[,mʌltɪ'plɪsəti]中文解析n. 多样性;[物] 多重性词laissez-faire音标[ˌleseˈfɛr]中文解析adj. [法]放任主义的,自由放任的;=“laisser-faire”n. 自由放任主义,无干涉主义词mercantile 音标[ˈmə:kənˌtil, -ˌtaɪl, -tɪl] 中文解析adj. 贸易的,商业的,商人的;重商主义的词peculiar音标[pɪ'kjulɪɚ]中文解析adj. 特殊的;独特的;奇怪的;罕见的n. 特权;特有财产词cement 音标[sə'mɛnt]中文解析vt. 巩固,加强;用水泥涂;接合vi. 粘牢n. 水泥;接合剂词entrepreneurial 音标[,ɔntrəprə'nə:riəl] 中文解析adj. 创业的,具有企业精神的;企业性质的词pastoralist音标['pæstərəlist]中文解析n. 田园诗(或曲、剧、画等)的作者词pasture音标['pɑ:stʃə, 'pæs-]中文解析n. 草地;牧场;牧草vt. 放牧;吃草词self-assertion 音标['selfə'sə:ʃən]中文解析n. 自信;自作主张;一意孤行词civilization音标[,sivilai'zeiʃən, -li'z-]中文解析n. 文明;文化词literally 音标['litərəli]中文解析adv. 照字面地;逐字地词intrinsically 音标[in'trinsikəli]中文解析adv. 本质地;内在地;固有地词fragile音标['frædʒail]中文解析adj. 脆的;易碎的词seafarer音标['si:fεərə]中文解析n. 船员;航海家词circumvent音标[,sə:kəm'vent]中文解析vt. 包围;陷害;绕行词ecology音标[i:'kɔlədʒi]中文解析n. 生态学;社会生态学词predator音标['predətə]中文解析n. [动] 捕食者;[动] 食肉动物;掠夺者词barren音标['bærən]中文解析adj. 贫瘠的;不生育的;无益的;沉闷无趣的;空洞的n. 荒地词insecure音标[,insi'kujə]中文解析adj. 不安全的;不稳定的;不牢靠的词monetary音标['mʌnitəri]中文解析adj. 货币的;财政的TPO16-2 Development of the Periodic Table词periodic音标[,piəri'ɔdik]中文解析adj. 周期的;定期的词recurrence音标[riˈkʌrəns]中文解析n.复回,重现;反复,隐现;[数]循环;重新提起词atomic 音标[əˈtɑmɪk]中文解析adj.原子的;原子能的,原子武器的;极微的词proton音标[ˈproˌtɑn]中文解析n. [物]质子词monumental音标[ˌmɑnjəˈmɛntl:]中文解析adj.不朽的;纪念碑的;非常的词interplay 音标[ˈɪntɚˌple]中文解析n.相互作用词successive音标[səkˈsɛsɪv]中文解析adj.连续的,相继的;继承的,接替的;逐次词interval音标[ˈɪntəvəl]中文解析n.间隔;幕间休息;(数学)区间词horizontal音标[ˌhɔrɪˈzɑntl]中文解析adj.水平的,卧式的;地平线的;[植](枝条)平层的;同一行业的,同阶层的n.水平线;水平面;水平位置;水平的物体词automatically音标[ˌɔtəˈmætɪkl:ɪ]中文解析adv.自动地;无意识地;不自觉地;机械地词forerunner 音标['fɔ:,rʌnə]中文解析n. 先驱;先驱者;预兆词sufficient音标[sə'fiʃənt]中文解析adj. 足够的;充分的词farsighted音标中文解析有远见的,目光如炬的,有先见之明的词Indium音标[ˈindiəm]中文解析n.铟词arsenic音标[ˈɑ:sənɪk]中文解析n.砷;三氧化二砷,砒霜adj.砷的,含砷(主要指五价砷)的词selenium音标[英] [siˈli:niəm][美] [sɪˈliniəm]中文解析n.硒词cadmium音标[英] [ˈkædmiəm][美] [ˈkædmiəm]中文解析n.镉词tin音标英] [tin][美] [tɪn]中文解析n.锡;罐头盒;马口铁;镀锡薄钢板adj.锡制的;假冒的;无价值的;蹩脚的vt.镀锡,包锡;给……包马口铁;包白铁词subsequent 音标[英] [ˈsʌbsikwənt][美] [ˈsʌbsɪˌkwɛnt,-kwənt]中文解析adj.后来的;随后的;作为结果而发生的;附随的词iodine音标英] [ˈaɪəˌdaɪn, -dɪn,-ˌdi:n][美] [ˈaɪəˌdaɪn, -dɪn, -ˌdin]中文解析n.<化>碘词isotope音标[英] [ˈaɪsəˌtəʊp][美] [ˈaɪsəˌtop]中文解析n.[化]同位素词neutron音标[英] [ˈnu:ˌtrɔn, ˈnju:-][美] [ˈnuˌtrɑn,ˈnju-]中文解析n.[物]中子词aluminum音标[英] [əˈlu:mənəm][美] [əˈlumənəm]中文解析n.<美>铝词designate音标[英] [ˈdeziɡneit][美] [ˈdɛzɪɡˌnet]中文解析vt.指明,指出;指派;表明,意味着;把……定名为adj.指定而尚未上任的;选出而尚未上任的词gallium音标[英] [ˈɡæliəm][美] [ˈɡæliəm]中文解析n.镓词correspondence音标英] [ˌkɔrisˈpɔndəns][美] [ˌkɔrɪˈspɑndəns, ˌkɑr-]中文解析n.一致,符合;对应;通信,信件; [艺]通感词gallium 音标[英] [ˈɡæliəm][美] [ˈɡæliəm]中文解析n.镓词eka-aluminum音标中文解析词germanium音标中文解析锗词gaseous 音标[英] [ˈgæsi:əs, ˈgæʃəs][美] [ˈɡæsiəs,ˈɡæʃəs]中文解析adj.气态的,似气体的;无实质的;〈美俚〉不可靠的;气性词residual音标[英] [rɪˈzɪdʒu:əl][美] [rɪˈzɪdʒuəl]中文解析adj.残余的;残留的n.剩余;残渣词nitrogen音标[英] [ˈnaitrədʒən][美] [ˈnaɪtrədʒən]中文解析n.[化]氮,氮气词isolate音标[英] [ˈaisəleit][美] [ˈaɪsəˌlet]中文解析vt.使隔离,使孤立; [电]使绝缘; [化]使离析; [微]使细菌分离vi.隔离,孤立n.[微]分离菌;隔离种群adj.隔离的,分离的;孤立的词argon音标[英] [ˈɑ:ɡɔn][美] [ˈɑrˌɡɑn]中文解析n.氩词helium音标[英] [ˈhi:li:əm][美] [ˈhiliəm]中文解析n.<化>氦词spectrum音标[英] [ˈspektrəm][美] [ˈspɛktrəm]中文解析n.[物理学]谱,光谱:辐射源,能谱;光谱相片;范围;系列,范围,幅度词postulate音标[英] [ˈpɔstʃəˌleɪt][美] [ˈpɑstʃəˌlet]中文解析vt.假定;提出要求;视……为理所当然n.假定;先决条件;基本原理词neon音标[英] [ni:ən][美] [ˈniˌɑn]中文解析n.<化>氖;霓虹灯词krypton音标[英] [ˈkriptɔn][美] [ˈkrɪpˌtɑn]中文解析n.氪词xenon音标[英] [ˈzenɔn][美] [ˈziˌnɑn]中文解析n.<化>氙TPO16-3 Planets in Our Solar System词hub 音标[英] [hʌb][美] [hʌb]中文解析n.轮轴;中心,焦点;(电器面板上的)电线插孔; [计]集线器词rotating音标[rəuˈteitɪŋ]中文解析v.(使某物)旋转[转动](rotate的现在分词);(使某人或某物)轮流[按顺序循环]词numerous音标[英] [ˈnju:mərəs][美] [ˈnumərəs,ˈnju-]中文解析adj.很多的,许多的;数量庞大的数量庞大的;数不清的词asteroid音标[英] [ˈæstəˌrɔɪd][美] [ˈæstəˌrɔɪd]中文解析n.[天]小行星;海盘车;海星adj.星状的词comet音标[英] [ˈkɔmit][美] [ˈkɑmɪt]中文解析n.[天]彗星;孛词meteoroid音标[ˈmi:tjərɔid]中文解析n.流星体词estimated音标['estimetid]中文解析adj.估计的;预计的;估算的v.估计,评价,评估(estimate的过去式和过去分词);(粗略)估计(…的距离、价值、数目、大小、重量、费用等),估量,预算词gravitational音标[ˌɡræviˈteiʃnəl]中文解析adj.万有引力的,重力的;地心吸力的词elliptical 音标[iˈliptikəl]中文解析adj.椭圆的;像椭圆形的;省略的词orbit 音标[英] [ˈɔ:bit][美] [ˈɔrbɪt]中文解析n.轨道;势力范围;眼眶;(人生的)旅程,生活过程vt.& vi.在……轨道上运行,环绕轨道运行vi.盘旋;绕轨道运行vt.绕轨道而行;进入轨道词terrestrial 音标[英] [təˈrestri:əl][美] [təˈrɛstriəl]中文解析adj.陆地的;地球的;人间的; <天>类地行星的n.地球人,陆地生物词diameter音标[英] [daiˈæmitə][美] [daɪˈæmɪtɚ]中文解析n.直径,直径长;放大率词correlation音标[英] [ˌkɔ:rəˈleɪʃən,ˌkɔr-][美] [ˌkɔrəˈleʃən, ˌkɑr-]中文解析n.相互关系;相关性词dimension音标[英] [diˈmenʃən][美] [dɪˈmɛnʃən,daɪ-]中文解析n.尺寸; [复]面积,范围; [物]量纲; [数]次元,度,维adj.(石料,木材)切成特定尺寸的vt.把……刨成(或削成)所需尺寸;标出尺寸词density音标[英] [ˈdensiti][美] [ˈdɛnsɪti]中文解析n.密度;稠密,浓厚; [物]浓度,比重;愚钝词composition音标[英] [ˌkɔmpəˈziʃən][美] [ˌkɑmpəˈzɪʃən]中文解析n.作文,作曲;创作;构图,布置;妥协,和解词dense音标[英] [dens][美] [dɛns]中文解析adj.密集的,稠密的;浓密的,浓厚的;愚钝的词rocky音标[英] [ˈrɔki:][美] [ˈrɑki]中文解析adj.多岩石的;坚如磐石的,坚硬的;麻木的;头晕目眩的词metallic音标[英] [miˈtælik][美] [məˈtælɪk]中文解析adj.金属的;金属性的;金属制的;含金属的词percentage 音标[英] [pəˈsentidʒ][美] [pɚˈsɛntɪdʒ]中文解析n.百分比,百分率;比例,部分; [数]百分法; [商]手续费词hydrogen音标[英] [ˈhaidrədʒən][美] [ˈhaɪdrədʒən]中文解析n.<化>氢词helium音标[英] [ˈhi:li:əm][美] [ˈhiliəm]中文解析n.<化>氦词methane音标[英] [ˈmeθˌeɪn][美] [ˈmɛθˌen]中文解析n.<化>甲烷,沼气词ammonia音标[英] [əˈməʊnjə][美] [əˈmonjə]中文解析n.氨;氨水;氨气词molecule 音标[英] [ˈmɔlikju:l][美] [ˈmɑlɪˌk jul]中文解析n.分子;微小颗粒词evaporate音标[英] [iˈvæpəreit][美] [ɪˈvæpəˌret]中文解析vt.& vi.(使某物)蒸发掉vi.消失;发散气体;蒸发vt.使脱水;使蒸发;使挥发;使沉淀词velocity音标[英] [viˈlɔsiti][美] [vəˈlɑsɪti]中文解析n.速率,速度;周转率;高速,快速词mass音标[英] [mæs][美] [mæs]中文解析n.大量,大多;块,堆,团; [物理学]质量;弥撒曲vt.& vi.(使)集中,聚集adj.群众的;大规模的;整个的;集中的vi.聚集起来vt.使集合词molecular音标[英] [məˈlekjələ][美] [məˈlɛkjəlɚ]中文解析adj.分子的,由分子组成的词gravity音标[英] [ˈɡræviti][美] [ˈɡrævɪti]中文解析n.重力;万有引力,地心引力;重要性,严重性;严肃,庄重词carbon音标[英] [ˈkɑ:bən][美] [ˈkɑrbən]中文解析n.[化学]碳;(一张)复写纸; [电]碳精棒[片,粉],碳精电极;复写的副本adj.碳的;碳处理的词dioxide音标[英] [daɪˈɔksaɪd][美] [daɪˈɑksaɪd]中文解析n.[化]二氧化物词infinitesimally音标中文解析adv.无限小地词portion音标[英] [ˈpɔ:ʃən][美] [ˈpɔrʃən, ˈpor-]中文解析n.一部分;一份遗产(或赠与的财产);嫁妆;分得的财产vt.把……分成份额;分配;把……分给(to);命运注定词astronomer音标[英] [əˈstrɔnəmə][美] [əˈstrɑnəmɚ]中文解析n.天文学者,天文学家词essentially音标[iˈsenʃəli]中文解析adv.本质上,根本上;本来;“essential“的派生词hypothesize音标[英] [haiˈpɔθisaiz][美] [haɪˈpɑθɪˌsaɪz]中文解析v.假设,假定,猜测词primordial音标[praɪˈmɔrdiəl]中文解析adj.初生的,初发的,原始的词condense音标[英] [kənˈdens][美] [kənˈdɛns]中文解析vt.& vi.(使)变稠或变浓,浓缩;(使)凝结;精简;液化vt.变浓缩;使更紧密词void音标[英] [vɔid][美] [vɔɪd]中文解析adj.空的,空虚的,没人住的;(职位)空缺着的;无效的n.太空,宇宙空间;空位,空隙;空虚感,寂寞的心情vt.使无效;宣布……作废;取消;排泄。
Oxford5000单词表

Oxford5000单词表AIDSabolishabortionabsenceabsentabsorbabstractabsurdabundanceabuseacademyaccelerateaccentacceptanceaccessibleaccidentallyaccommodateaccomplishaccomplishmentaccordanceaccordinglyaccountabilityaccountableaccountantaccountingaccumulateaccumulationaccuracyaccuratelyaccusationaccusedacidacquisitionacreactivateactivationactivistacuteadaptationaddictionadditionallyadequateadequatelyadhereadjacentadjustadjustmentadministeradministrativeadministratoradmissionadolescentadoptionadverseadvocateagricultural agriculture aidealbeitalertalienalign alignment alike allegation allege allegedly alliance allocate allocation allowance ally alongside altogether aluminium amateur ambassador ambulance amend amendment amid amusing analogy analyst ancestor anchor angel animation annually anonymous anticipate anxiety apology apparatus appealing appetite applaud applicable applicant appoint appreciation appropriately arbitrary architectural archive arena arguably armarrayarrow articulate artworkassemble assembly assert assertion asset assign assistance assumption assurance assure astonishing asylum atrocity attachment attain attendance attorney attribute auction audioaudit authentic authorize auto automatic automatically autonomy availability await awareness awkward backdrop backing backup badgebail balanced ballet balloon ballot banner barebarely bargain barrel basement basket bassbat battlefield baybeam beast behalf beloved benchbiasbidbind biography biological bishop bizarre blade blanket blastbleed blendbless blessing blowboastbold bombing bonus booking boom boost bounce bound boundary bow breach breakdown breakthrough breedbrick briefly broadband broadcaster broadly browser brutalbuck buddy bufferbugbulk burden bureaucracy burialburstcabin cabinet calculation canal candle canvas capability capitalism capitalistcatalogue catercattle caution cautious cavecease cemetery certainty certificate challenging chamber championship chaos characterize charm charming charter chase cheek cheerchoirchop chronic chunk circuit circulate circulation citizenship civiccivilian civilization clarify clarityclash classification classify cleaning clerkcliffclingclinic clinicalclipclosure cluster coalition coastal cocktail cognitive coincide coincidence collaborate collaboration collectivecomic commander commence commentary commentator commerce commissioner commodity communist companion comparable comparative compassion compel compelling compensate compensation competence competent compile complement completion complexity compliance complication comply compose composer composition compound comprehensive comprise compromise compulsory compute conceal concede conceive conception concession concrete condemn confer confess confession configuration confine confirmation confront confrontation confusion congratulate congregation congressionalconsequently conservation conserve considerable considerably consistency consistently consolidate conspiracy constituency constitute constitution constitutional constraint consult consultant consultation consumption contemplate contempt contend contender content contention continually contractor contradiction contrary contributor controversial controversy convenience convention conventional conversion convey convict conviction convincing cooperate cooperative coordinate coordination coordinator copcopecopper copyright corporation correction correlate correlation correspond correspondence correspondentcounselling counsellor counter counterpart countless coup courtesy coverage crackcraftcrawl creativity creator credibility crediblecreep critically critiquecrowncrudecruisecrushcrystalcuecultcultivate curiosity curious curriculum custodycutecuttingcynicaldairydam damaging dare darkness database dawndead deadline deadlydealerdebrisdebut decision-making decisivedeck declaration dedicated dedication deeddeemdefaultdefectdelete delicate democracy democratic demon demonstration denial denounce dense density depart dependence dependent depict deploy deployment deposit depression deprive deputy derive descend descent designate desirable desktop desperately destruction destructive detain detection detention deteriorate determination devastate devildevise devote diagnose diagnosis dictate dictator differ differentiate dignity dilemma dimension diminishdip diplomat diplomatic directory disability disabled disagreement disappointdiscourage discourse discretion discrimination dismissal disorder displace disposal dispose dispute disrupt disruption dissolve distant distinct distinction distinctive distinguish distort distract distress disturb disturbing divediverse diversity divertdivine divorce doctrine documentation domain dominance dominant donation donordosedot downtown drain dramatically driftdriving drought drowndualdubdulldumbdumpduo duration dynamic eager earnings easeeffectiveness efficiency efficientlyego elaborate elbow electoral electronics elegant elementary elevate eligible eliminateeliteembark embarrassment embassy embed embody embrace emergence emission emotionally empire empirical empower enact encompass encouragement encouraging endeavour endless endorse endorsement endure enforce enforcement engagement engaging enjoyable enquire enrichenrolensue enterprise entertaining enthusiast entitleentityen trepreneur envelope epidemic equality equation equipessence essentially establishment eternalethicethnic evacuate evaluation evident evoke evolution evolutionary evolve exaggerate exceed excellence exception exceptional excess excessive exclude exclusion exclusive exclusively execute execution exert exhibitexileexitexotic expansion expenditure experimental expertise expire explicit explicitly exploit exploitation explosive exposure extension extensive extensively extract extremist fabric fabulous facilitate faction facultyfadefailed fairness fakefeat federal feeding feminist feverfibre fiercefilm-maker filterfine firearm firefighter firework firmfirmlyfitfixtureflat flavour flaw flawed fleefleetflesh flexibility flourish fluidfondfool footage forbid forecast foreigner forge format formation formerly formula formulate forth forthcoming fortunate forum fossil foster foundation founder fraction fragile fragment framework franchise frankly fraud freely frequentfundamentally fundraising funeral furious gallon gambling gaming gathering gaygazegear gender gene generic genetic genius genocide genuine genuinely gesturegigglance glimpse globalization globe glorious glory golden goodness gorgeous governance governor grace graphic graphics graspgrave gravity greatly greenhouse gridgriefgringrindgrip grocery guerrilla guidance guideline guiltguthabitathail halfwayhalt handfulharvest hatredhaunt hazard headquarters heal healthcare heighten helmet henceherb heritage hidden hierarchy high-profile highway hilarioushinthiphistorian homeland homeless honestyhook hopeful hopefully horizonhorn hostage hostile hostility humanitarian humanity humble hunger hydrogen hypothesis icon identical identification ideological ideologyidiot ignorance illusion imagery immense immigration imminent immune implement implementation implication imprison imprisonmentincorporate incorrect incur independence index indication indicator indictment indigenous induce indulge inequality inevitable infamous infantinfectinfer inflation inflict influential info infrastructure inhabitant inherent inherit inhibit initiateinject injection injusticeinkinmate innovation innovative inputinsert insertion insider inspect inspection inspector inspiration installation instant instantly instinct institutional instruct instrumental insufficient insultintactintake integral integrateintent interact interaction interactive interface interfere interference interim interior intermediate interpretation interval intervene intervention intimate intriguing invade invasion investigator investor invisible invoke involvement ironic ironically irony irrelevant isolate isolated isolationjailjetjoint journalism judicial junction jurisdiction juryjust justification kidnap kidney kingdomkitladladder landing landlord landmark lanelaplarge-scale laserlatelyleakleaplegacy legend legendary legislation legislative legislature legitimate lengthylens lesbian lesser lethalliable liberal liberation liberty license lifelong lifetime lighting likelihood likewise limb limitation line-up linearlinger listing literacy literally literarylitrelitterliverlobbyloglogiclogolong-standing long-time loomlooplotteryloyal loyaltylyric machinery magical magistrate magnetic magnificent magnitude mainland mainstreammandatory manifest manipulate manipulation manufacture manufacturing manuscript marathon march margin marginal marine marker marketplace martial mask massacre mate mathematical mature maximize mayor meaningful meantime mechanic mechanical mechanism medal medication medieval meditation melody membership memo memoir memorable memorial mentor merchant mercymere merely merge merger merit metaphor methodology midst migration militant militiamillminer minimal minimize mining ministrymissile mob mobility mobilize mode moderate modest modification momentum monk monopoly monster monthly monument morality moreover mortgage mosque motion motivate motivation motive motorist moving municipal mutual myth naked namely nasty nationwide naval navigation nearby necessity neglect negotiate negotiation neighbouring nestnetneutral newly newsletter niche noblenod nominate nomination nominee non-profit nonetheless nonsense noonnorm notable notablynursery nursing nutrition obesity objection oblige observer obsess obsession obstacle occasional occupation occupy occurrence odds offender offering offspring ongoing openly opera operational operatoroptoptical optimism optimistic oral orchestra organic organizational orientation originate outbreak outfitoutingoutlet outlook output outrage outsider outstanding overcome overlook overly overnight overseas oversee overturn overwhelm overwhelming ownership oxygen packetpadparliamentary part-time partial partially participation partnership passing passionate passive password pastor patch patent pathway patience patrol patron pausepeak peasant peculiar peer penalty perceive perception permanently persist persistent personnel petition philosopher philosophical physician pillpioneer pipeline piratepitpity placement pleaplead pledgeplug plungepolepollpondpop portfolio portion portraypost-war postpone potentiallypredator predecessor predictable predominantly preference pregnancy prejudice preliminary premier premise premiumprescribe prescription presently preservation preside presidency presidential prestigious presumably presume prevail prevalence prevention preypride primarily principal prior privatization privilege probability probable probe problematic proceed proceedings proceeds processing processor proclaim productive productivity profitable profound programming progressive prohibit projection prominent promising promotion promptprospective prosperity protective protein protester protocol province provincial provision provoke psychiatric psychological publicity publishing pulsepump punchpunkpurely pursuit puzzle queryquest questionnaire quotaracial racism racistradar radiation radicalrageraidrailrally random ranking raperatratingratio rationalrayreadily realization realmrear reasonably reasoning reassure rebel rebellion rebuild receiverrecruitment referee referendum reflection reform refuge refugee refusal regain regardless regime registration regulate regulator regulatory rehabilitation reign reinforce rejection relevance reliability relieve relieved reluctant remainder remains remarkable remarkably remedy reminder removal render renew renowned rental replacement reportedly reporting representation reproduce reproduction republic resemble reside residence residential residue resign resignation resistance resolution respective respectively restorationretrieve revelation revenge revenue reverse revision revival revive revolutionary rhetoric ridiculous rifleriotripriskyritualrivalrob robbery robustrockrocketrod romance roserotate rotation roughlyruinruling rumour sack sacred sacrifice saintsake sanction satisfaction say scandal scare scattered scenario sceptical scholar scholarship scope scratch screening screw scrutiny seal secular seekersensitivity sentiment separation serialset-up settlement settler severely sexuality sexyshaped shareholder shattersheer shipping shocking shootshoreshort-term shortage shortly shrinkshrugsiblingsigh signature significance simulate simulation simultaneously sinsituated sketch skilledskipskullslamslapslash slavery sloganslotsmashsnapso-called soaksoar socialistsolesolely solicitor solidaritysolo somehowspark specialize specialized specification specify specimen spectacle spectacular spectator spectrum speculate speculation spellspherespillspinspinespitespoil spokesman spokesperson spokeswoman sponsorship sporting spotlight spousespysquad squeezestab stability stabilize stakestallstance standing starkstarve statistical steadily steamsteerstem stereotype stimulate stimulusstirstorage straightforward strainstrand strategic strengthen strictlystructural stumble stun stunning submission subscriber subscription subsequent subsequently subsidy substantial substantially substitute substitution subtle suburb suburban succession successive successor sucksue suffering sufficient sufficiently suicide suite summit super superb superior supervise supervision supervisor supplement supportive supposedly suppress supreme surge surgeon surgical surplus surrender surveillance survival survivor suspend suspension suspicion suspicious sustain sustainable swallow swing sword symbolictactictacticaltagtaptaxpayer technological teenstemple temporarily tempttenant tendency tendertensiontenureterminal terminatetermsterrainterriblyterrificterrifyterritoryterrorterrorism terroristtestify testimony testing textbook texture thankfully theatricalthefttheology theoretical therapist thereafter therebythesis thorough thoroughly thought-provoking thoughtful thread threshold thrilledthrivethumbtidetightentimbertimelytimingtissuetobacco tolerancetoptorturetosstotal tournament toxictrace trademark trading tragedy tragictrailtrailertrait transaction transcript transformation transit transmission transmit transparency transparent transportation traptrauma treasure treaty tremendous tribaltribetribunal tribute triggertrilliontriotriumphtrooptrophy troubled trustee tsunami tuition turnout turnovertwist ultimate unacceptable uncertainty undergo undergraduate underlying undermine undertake undoubtedly unfold unfortunate unifyuniteunity universal unprecedentedupcoming upgrade uphold urgent usage useless utility utilize utterly vacuum vague valid validity vanish variable variation varied vein venture verbal verdict verify verse versus vertical vessel veteran viable vibrant vice vicious viewpoint villager violate violation virtuevisa visible vocal voluntary votingvow vulnerability vulnerable wander ward warehouse warfare warming warrant warrior weaken weave weed weekly weird welfare wellwell-being whatsoever wheatwhilst whip whoever wholly widen widespread widow width willingness wipe wisdom wit withdraw withdrawal workforce workout workplace workshop worm worship worthwhile worthy wristyellyield youngster。
星火2013考研英语词汇词根+联想+图解记忆法Unit 11-15

kin n.家属,亲属
liability n.责任,义务;债务;
惹麻烦的人(或物)
liable a.有义务的,有责任的;
有倾向的,易于…的,可能做…
likelihood n.可能性
manufacture vt.制造,加工
n.制造,制造业;产品,工业品
counter 2 vt.反对,反击
a./ad.相反的(地)
dash v./n.猛冲,突进;猛掷 n.破折号
display vt./n.陈列,展览 ;显示
division n.分,分割;分配,分担;
除(法);部门,科,处
engage v.(使)参加,(使)从事于;
吸引(注意力、兴趣等);
awe n.敬畏,惊叹
vt.使敬畏,使惊叹
awful a.极坏的,讨厌的;
威严的,可怕的;非常的,很多的
bare a.赤裸的,光秃的,空的;
仅有的,最基本的 v.露出,暴露
barely ad.仅仅,勉强,几乎没有;
赤裸裸地,无遮蔽地
consist vi.由…组成,由…构成;
在于,存在于;一致,符合
balance v.使平衡
n.天平,秤;平衡,均衡;差额,结存,余款
concept n.概念,观念,思想
conception n.观念,概念;构想,设想;怀孕
deceit n.欺骗(行为)
explore v.探测,勘探;探索,探究
explosive n.炸药,爆炸物
a.爆炸(性)的,易爆发的;(性情等)暴躁的
minority n.少数,少数派;少数民族
00015 英语二 完整单词 词库 可导入欧陆词典手机学习

book booklet bookmark bookshelf boom boost boot booth border bore boring born borrow boss both bottle bottom bound boundary bow bowel bowl box boxing boy brain brake branch brand brave bread breadth break breakfast breath breathe breed breeze bribe brick bride bridegroom bridge brief bright brilliant
beast beat beautiful beauty because become bed bedroom bee beef beehive beer before beg beggar begin beginning behalf behave behavio(u)r behind being belief believe bell belong belongings beloved below belt bench bend beneath beneficial benefit beside besides best best-seller bet betray better between beyond bible
acquaint acquaintance acquire acquisition acre across act action active activity actor actress actual acute AD adapt add addition additional address adequate adjust administration admire admission admit adopt adult advance advanced advantage adventure advertise advertisement advice advisable advise advocate affair affect affection afford afraid Africa African
姿态估计算法汇总基于RGB、RGB-D以及点云数据

姿态估计算法汇总基于RGB、RGB-D以及点云数据作者:Tom Hardy点击上⽅“3D视觉⼯坊”,选择“星标”⼲货第⼀时间送达作者⼁Tom Hardy@知乎编辑⼁3D视觉⼯坊姿态估计算法汇总|基于RGB、RGB-D以及点云数据主要有整体⽅式、霍夫投票⽅式、Keypoint-based⽅式、Dense Correspondence⽅式等。
实现⽅法:传统⽅法、深度学习⽅式。
数据不同:RGB、RGB-D、点云数据等;标注⼯具实现⽅式不同整体⽅式整体⽅法直接估计给定图像中物体的三维位置和⽅向。
经典的基于模板的⽅法构造刚性模板并扫描图像以计算最佳匹配姿态。
这种⼿⼯制作的模板对集群场景不太可靠。
最近,⼈们提出了⼀些基于深度神经⽹络的⽅法来直接回归相机或物体的6D姿态。
然⽽,旋转空间的⾮线性使得数据驱动的DNN难以学习和推⼴。
1.Discriminative mixture-of-templates for viewpoint classification2.Gradient response maps for realtime detection of textureless objects.paring images using the hausdorff distance4.Implicit 3d orientation learning for 6d object detection from rgb images.5.Instance- and Category-level 6D Object Pose Estimation基于模型2.Deep model-based 6d pose refinement in rgbKeypoint-based⽅式⽬前基于关键点的⽅法⾸先检测图像中物体的⼆维关键点,然后利⽤PnP算法估计6D姿态。
1.Surf: Speeded up robust features.2.Object recognition from local scaleinvariant features3.3d object modeling and recognition using local affine-invariant image descriptors and multi-view spatial constraints.5.Stacked hourglass networks for human pose estimation6.Making deep heatmaps robust to partial occlusions for 3d object pose estimation.7.Bb8: A scalable, accurate, robust to partial occlusion method for predicting the 3d poses of challenging objects without using depth8.Real-time seamless single shot 6d object pose prediction.9.Discovery of latent 3d keypoints via end-toend geometric reasoning.10.Pvnet: Pixel-wise voting network for 6dof pose estimation.Dense Correspondence/霍夫投票⽅式1.Independent object class detection using 3d feature maps.2.Depth encoded hough voting for joint object detection and shape recovery.3.aware object detection and pose estimation.4.Learning 6d object pose estimation using 3d object coordinates.5.Global hypothesis generation for 6d object pose estimation.6.Deep learning of local rgb-d patches for 3d object detection and 6d pose estimation.7.Cdpn: Coordinates-based disentangled pose network for real-time rgb-based 6-dof object pose estimation.8.Pix2pose: Pixel-wise coordinate regression of objects for 6d pose estimation.9.Normalized object coordinate space for categorylevel 6d object pose and size estimation.10.Recovering 6d object pose and predicting next-bestview in the crowd.基于分割深度学习相关⽅法1.PoseCNN: A convolutional neural network for 6d object pose estimation in cluttered scenes.2.Render for cnn: Viewpoint estimation in images using cnns trained with rendered 3d model views.6.Robust 6D Object Pose Estimation in Cluttered Scenesusing Semantic Segmentation and Pose Regression Networks - Arul Selvam Periyasamy, Max Schwarz, and Sven Behnke. [[Paper]数据格式不同根据数据格式的不同,⼜可分为基于RGB、RGB-D、点云数据的识别算法。
高中生物英文

高中生物英文篇一:高中必修生物中的英语词汇BiologyAaccelerate [?k'sel?reit] vt. & vi. (使)加快, (使)增速acid ['?sid] n.〈化〉酸acidic[?'s?d?k] adj.酸的,酸性的algae['?ld?i:] n. 水藻aluminum [?'lu:m?n?m] n.铝Alamino [?'mi:n??, '?m?,n??] adj.氨基的ammonia [?'m??nj?] n. 氨amoeba [?'mi:b?] n. 变形虫apparent [?'p?r?nt] adj. 显然的, 明白的, 清晰可见的; 表面上的, 貌似(真实)的aqueous ['eikwi?s] adj. 水的,水成的atom ['?t?m] n. 原子; 微粒, 微量Bbase n.碱basic adj.碱的,碱性的beryllium [b?'rilj?m] n铍Bebiomolecule生物分子bloodstream ['bl?d,stri:m] n. 血流,体内循环的血液bond [b?nd] n.联系, 关系; 连接, 接合, 结合boron ['b?:r?n] n. 硼BCcalcium ['k?lsi?m] n. 钙Cacapillary ['k?p?,leri:] n. 毛细管;毛细血管capture vt. 俘获;夺取, 占领carbohydrate [,kɑ:b?u'haidreit] 碳水化合物; 糖类carbon n. 碳Ccatalyst ['k?tl?st] n. <化>催化剂,触媒cell n. 〈生〉细胞cell wall细胞壁cellular['selj?l?] adj. 由许多小单元组成的;蜂窝状的;多孔的cellulose ['selj?,l??s, -,l??z] n. 细胞膜质;纤维素chemical behavior化学行为chemical formula化学式chemical equation化学方程式chemically bine化学结合chlorine ['kl?:ri:n] n. 氯Cl;氯气chlorophyll ['kl?:r?f?l, 'kl??r-] n. 叶绿素cholesterol [k?'lest?r?l] n. 胆固醇chromium ['kr??mi:?m] n. 铬Crchromosome ['kr??m?,s??m] n. <生>染色体cobalt [k?'b?:lt,'k?ub?:lt] n. 钴Cocollide [k?'laid] vi. 相撞, 碰撞collision [k?'li??n] n. 碰撞, 冲突, 抵触pose vt. 组成, 构成pound ['k?mpaund] n. 复合物, 化合物;vt. 使混合, 使合成pound light microscope复合光学显微镜condensation [,k?nden'sen, -d?n-] n. 冷凝,凝聚copper ['k?p?] n. 铜Cucoronary artery冠状动脉corn syrup玉米浆,淀粉糖浆correspondence [,k?ris'p?nd?ns] n. 一致, 相似covalent [k?u'veil?nt] adj. 共有原子价的,共价的covalent bond共价键critical ['kritik?l] adj. 决定性的, 关键性的, 危急的crosslink ['kr?sli?k] n. 交联,交键crystal ['krist?l] n. 结晶(体)Ddense adj. 密集的, 稠密的, 浓密的detectable adj. 可发觉的,可看穿的diagnose ['dai?ɡn?uz] vt. 诊断diagram ['dai?ɡr?m] n. 图解, 简图, 图表diffusion n. <物,化>扩散;漫射digestion [da?'d?est??n, d?-] n. 消化能力dimensional [d?'men??n?l] adj. <物>量纲的;<数>因次的;维的disaccharide [dai's?k?raid] n. 二糖dissolve [di'z?lv]vt. & vi. (使)溶解double bonds双键Eelectron n. 电子electron energy level电子层electron microscope电子显微镜element n.元素enzyme ['enza?m] n. <生化>酶eukaryote [ju'k?ri?ut] n. 真核细胞external [eks't?:nl] 外面的, 外部的Ffatty acid脂肪酸fluid mosaic model流动镶嵌模型; 膜的流动镶嵌模型fluorine n. 氟FGgas n. 气体gel n. 凝胶, 冻胶glycogen ['ɡlaik?ud?en] n. 肝糖,糖原质glucagon['ɡlu:k?ɡ?n] n. 胰高血糖素,胰增血糖素glucose ['glu:,k??s]n. 葡萄糖gradient n.道路的斜度, 坡度;变化程度grindHhigh-density lipoprotein(HDL)高密度脂蛋白homeostasis [,h?umi?u'steisis] n. 动态静止,动态平衡,(社会群体的)自我平衡,原状稳定hormone ['h?:,m??n] n. <生化>(刺激生长的)荷尔蒙,激素hydrogen n. 氢Hhydroxide [hai'dr?ksaid] n. 氢氧化物,羟化物hypotheses臆测,假定Iidentical adj. 同一的illustration [,?l?'strei??n] n. 说明, 图解, 图示immunity [?'mju:n?ti:] n. 免疫力infestation n. (害虫、盗贼等)群袭,出没,横行insulin ['?ns?l?n] n. 胰岛素intermolecular分子间的iodine ['a??,da?n, -d?n, -,di:n] n. 碘Iion n. <物>离子ionic bond 离子键iron n.铁Feirritant adj. 有刺激性的;使无效的, 使失效的isotope ['a?s?,t??p] n. <核>同位素Kkinetic [k?'net?k, ka?-] adj. <物>动力的,由运动引起的Llipoprotein .[,l?p??'pr??,ti:n, -ti:?n, ,la?p??-] n. <生化>脂蛋白liver n. 肝脏lonic band离子键low-density lipoprotein(LDL)低密度脂蛋白Mmagnesium [m?g'ni:zi:?m, -??m] n. 镁Mgmagnify['m?ɡnifai] vt. 放大; 扩大mammal n. 哺乳动物manganese ['m??g?,ni:z, -,ni:s] n. 锰Mnmaximum ['m?ksim?m] n. 最大的量、体积、强度等mechanism ['mek?niz?m] n. 机械装置metabolism [m?'t?b?,l?z?m] n. 新陈代谢microorganism [,ma?kr??'?:g?,n?z?m] n. 微生物 mineral n. 矿物; 矿石; 矿物质mixture混合物molecule ['m?likju:l] n. 分子molten adj. 熔化的;熔融的molybdenum [m?'libdin?m] n. 钼Momonosaccharide.[,m?n?u's?k?raid] n. 单糖multicellular [,m?lti'seljul?] adj. 多细胞的Nnerve impulse神经脉冲net charge净电荷neutral adj.<化>中性的neutron ['nu:,tr?n, 'nju:-] n. <物>中子nitrogen ['naitr?dn] n. 氮Nnitrogenous base含氮碱nonliving [n?n'livi?] adj.无生命的,缺乏生命nonmetal element非金属元素nuclei ['nju:kliai] n. 核,核心,原子核(nucleus的复数形) nucleic adj. 核的nucleolus [nju:'kli:?l?s] n. 核仁nucleotide ['nu:kli:?,ta?d, 'nju:-] n. <生化>核苷酸nucleus ['nju:kli?s] n. 中心, 核心O篇二:生物英文单词生物复习单词1要求会认的单词2、要求会背单词篇三:高中生物涉及的英文缩写高中生物涉及的英文缩写1.DNA:脱氧核糖核酸,是主要的遗传物质。
小学上册第七次英语第5单元寒假试卷

小学上册英语第5单元寒假试卷英语试题一、综合题(本题有100小题,每小题1分,共100分.每小题不选、错误,均不给分)1.The tortoise is slow but very _______ (聪明).2.How many strings does a typical guitar have?A. FourB. FiveC. SixD. SevenC3.________ (植物多样性改善) benefits ecosystems.4.What do we call the act of writing a letter?A. CorrespondenceB. CommunicationC. MessagingD. NotifyingA5.I can _______ (count) to twenty.6.The cockatoo has a fluffy _____ crest.7.I can dance ______.8.My favorite hobby is ______ (读书).9.My mom makes the best ________ (饼干) in the world.10.My mom makes _____ (晚餐) for us.11.My favorite movie is a ________ (冒险片).12.What is a synonym for "big"?A. SmallB. LargeC. TinyD. LittleB13.What is the main function of a compass?A. Measure temperatureB. Determine directionC. Calculate distanceD. Show timeB Determine direction14. A neutron star is incredibly ______ and dense.15.The __________ was a time when the world faced significant changes. (冷战)16.My dad loves to teach me about __________ (科学).17.We had fun making a video with our toy ____. (玩具名称)18.I write a diary every day to share my thoughts and ______ (感受). It helps me understand myself better.19.The __________ is a famous historical site.20.Did you see a _______ (小蓝鸟) in the garden?21.My dog loves to play fetch with a _________ (球).22.The ____ has big wings and can glide through the air.23.The __________ (历史的文化归属) influences perspectives.24.The __________ (历史的接受) varies across cultures.25.Which animal is known for its long neck?A. ElephantB. GiraffeC. DogD. HorseB26.What do we call the place where we learn?A. HomeB. SchoolC. ParkD. OfficeB27.I love to bake ______ during the holidays.28.The process of separating mixtures based on boiling points is called ______.29.The country known for its fjords is ________ (挪威).30.The _______ (老虎) is known for its strength.31.What do we call the process of testing a hypothesis in science?A. ExperimentationB. ObservationC. AnalysisD. ConclusionA32.I love to eat ___. (ice cream)33.What do we call a small insect that can produce silk?A. ButterflyB. SpiderC. CaterpillarD. Silkworm34.I like to build ______ (沙堡) at the beach with my friends.35.The __________ (历史的启迪) ignites passion.36.What is the name of the famous ancient city in Iraq?A. BabylonB. UrukC. NinevehD. All of the above37.What is the name of the famous painting by Leonardo da Vinci?A. The Starry NightB. Mona LisaC. The Last SupperD. Girl with a Pearl EarringB38.What do you use to brush your teeth?A. ToothbrushB. CombC. ToothpasteD. SoapA39.What is the capital of the Comoros?A. MoroniB. MoutsamoudouC. MitsamiouliD. DomoniA40.What do you call a person who studies education?A. EducatorB. TeacherC. ProfessorD. ScholarD41.What is the name of the famous wizarding school in Harry Potter?A. DurmstrangB. BeauxbatonsC. HogwartsD. Ilvermorny42.They are _______ (painting) the house.43.My sister loves to play with her _____.44.What is the opposite of shallow?A. DeepB. WideC. BroadD. All of the aboveA45.The __________ was a key moment in the history of democracy. (独立宣言)46.My brother loves to __________ (参加) community service.47.Magnetic fields are produced by moving ______.48.What do we call the first meal of the day?A. DinnerB. LunchC. BreakfastD. SnackC49.What do you call a story that is not true?A. FactB. FictionC. BiographyD. ReportB50.My grandmother loves __________ (收集) stamps.51.I want to ___ a musician. (become)52.What do bees make?A. MilkB. HoneyC. ButterD. JamB53.The park is ___. (fun)54.__________ (分子生物学) investigates the chemical foundations of biology.55.She has a new ________.56.Space is a near-perfect ______.ets are often described as "dirty snowballs" made of ice and ______.58.Which tree produces acorns?A. MapleB. PineC. OakD. BirchC59.The puppy is ______ in the sun. (sleeping)60.They are dancing _____ (快乐地).61.The ancient Greeks honored their deities through ________ and rituals.62.What is the capital of the Bahamas?A. NassauB. FreeportC. Marsh HarbourD. George Town63.The forecast says it might ______ (下雨) this evening.64.What do you use to cut paper?A. GlueB. ScissorsC. TapeD. Ruler65.The ancient civilizations of the Americas built ________ for religious ceremonies.66.The _______ of a pendulum is determined by its length.67.I saw a ________ flying in the air.68.What is a group of stars called?A. GalaxyB. UniverseC. Solar SystemD. ConstellationD69.I can travel with my ________ (玩具名称).70.Sedimentary rocks often contain ______, which are the remains of ancient plants and animals.71.The children are _____ in the sandbox. (playing)72.The chemical equation for photosynthesis shows the conversion of carbon dioxide and water into _______.73. A ____ is often seen chasing after butterflies and insects.74.My ______ is a talented actor.75.My sister is a _____ (演员) auditioning for roles.76.Which fruit is often used to make wine?A. GrapeB. AppleC. CherryD. PeachA77. A catapult uses tension to launch a ______.78.My friend gave me a ______ (拼图) for my birthday. It has many pieces and is very ______ (有趣的) to do.79.I want to _______ (学习)更多语言.80.My favorite way to spend a rainy day is ______.81.The first successful vaccine for smallpox was developed by ________.82.What do you call the process of plants making food?A. DigestionB. PhotosynthesisC. RespirationD. FermentationB83.The starfish can be found in ______ (海洋) around the world.84.I can ________ (solve) this math problem.85.What do we call the act of working together toward a common goal?A. CollaborationB. TeamworkC. CooperationD. All of the AboveD86.The game is ___. (exciting)87.The Earth's crust is subject to various geological ______.88.I enjoy building with my ________ (积木).89.The ________ was a significant battle during the American Revolutionary War.90. A solution's concentration can be expressed in ______.91.Environmental changes can impact the Earth's ______.92.The __________ is a famous mountain range in Europe.93.Which part of the plant grows underground?A. StemB. LeafC. RootD. Flower94. A __________ day is great for a family outing. (欢快的)95.The country known for its bamboo is ________ (中国).96.The _____ (狮子) is a symbol of strength and bravery.97.What is the term for the distance around a circle?A. DiameterB. RadiusC. CircumferenceD. AreaC98. A chemical reaction can produce _____ as a by-product.99.What is the main ingredient in sushi?A. RiceB. NoodlesC. BreadD. PotatoA Rice 100.My uncle builds ____ (houses) for a living.。
数学专业英语词汇(D)

数学专业英语词汇(D)d integrable d可积d integral d积分d'alembert principle 达朗贝尔原理d'alembert ratio test 达朗贝尔比例试验法d'alembert solution 达朗贝尔解d'alembertian 达朗伯符;达郎贝尔算子damped harmonic oscillation 阻尼谐振动damped oscillation 阻尼振动damped vibration 阻尼振动damping 阻尼damping factor 阻尼因子dantzig van de panne method 但泽范德潘方法darboux tangent 达布切线darboux theorem 达布定理data 数据data processing 数据处理data storage 数据存储器data storage register 数据存储寄存器death process 死亡过程death rate 死亡率debugging 堤序deca 十decade 十个decade scaler 十进制计数器decagon 十边形decahedron 十面体decameter 十米decay curve 衰变曲线deci 分decidability 可判定性decile 十分位数decimal 十进位的decimal arithmetic 十进算术decimal binary conversion 十二进制变换decimal digit 十进制数字decimal expansion 十进制展开decimal fraction 十进小数decimal notation 十进制记数法decimal number 十进小数decimal number system 十进制decimal of many places 多位十进小数decimal part 小数部分decimal place 小数位decimal point 小数点decimal representation 十进制记数法decimal system 十进制decimal to binary conversion 十二进制变换decimetre 分米decision 判定decision domain 决策域decision function 判定函数decision problem 判定问题decision procedure 判定过程decision space 判定空间decision theory 决策论decision variable 决策变量decision vector 决策向量decisive 决定的declination 倾斜decoder 译码器decomposability 可分解性decomposable form 可分解形式decomposable matrix 可分解矩阵decomposable operator 可分解算子decompose 分解decomposition 分解decomposition field 分解域decomposition formula 分解公式decomposition group 分解群decomposition in a direct sum 直和分解decomposition into linear factors 线性因子分解decomposition into partial fractions 部分分数分解decomposition operator 分解算子decomposition principle 分解原理decomposition theorem 分解定理decrease 减少decreasing function 递减函数decrement 减量dedekind axiom 绰金公理dedekind completion 绰金完备化dedekind cut 绰金切断dedekind domain 绰金环dedekind ring 绰金环dedekind set 绰金集dedekind sum 绰金和deduce 演绎deducibility 可推断deduction 演绎法deductive method 演绎法deductive proof 演绎证明defect 靠defect indices 扛数defect of operators 算子的靠defect of spline 样条的筐defect relation 控系defect subspaces 坑空间defective number 靠defective value 康deferent 圆心轨迹deficiency 靠deficiency index 扛标deficient number 靠definability 可定义性definable 可定义的define 定义definiendum 被定义者definiens 定义者defining contrast 定义对比defining equation 定义方程defining field 定义域defining relations 定义关系definite 定的definite divergence 定发散definite integral 定积分definiteness 梅性definition by induction 用归纳法定义definition by transfinite induction 依超限归纳法的定义deflation 降阶deform 使变形deformable 可变形的deformation 变形deformation ratio 形变比率deformation retract 形变收缩核deformation retraction 形变收缩degeneracy 退化degeneracy operator 退化算子degenerate 退化degenerate case 退化情况degenerate core 简并核degenerate distribution 退化分布degenerate eigenvalue 退化本盏degenerate extreme point 退化极值点degenerate kernel 退化核degenerate parabolic equation 退化抛物型方程degenerate polyhedron 退化多面体degenerate set 退化集degenerate simplex 退化单形degeneration 退化degree 次数degree of a polynomial 多项式的次数degree of a representation 表示度degree of accuracy 精确度degree of an equation 方程式的次数degree of approximation 近似度degree of freedom 自由度degree of inseparability 不可分次数degree of mapping 映射度degree of stability 稳定度degree of symmetry 对称度del 倒三角形del operator 倒三角形delay 延迟delay equation 延滞方程delay line store 延迟线存储器delay time 延迟时间delete 删去deleted neighborhood 去心邻域deletion 删除delocalization 非局部化delta function 狄垃克函数deltoid 形曲线demarcation 划分界线demi continuous 半连续的demonstrate 证明论证demonstration 证明denominate number 庚denomination 名称denominator 分母denote 指示dense 稠密的dense in itself 自密的dense in itself set 自密集dense set 稠集dense subset 稠子集denseness 稠密性denseness of set 集的密度densimetry 密度测定density 密度density distribution 密度分布density function 密度函数density matrix 密度矩阵density of distribution 分布密度density of simultaneous distribution 联合分布密度density theorem 密度定理denumerability 可数性denumerable 可数的denumerable set 可数集denumeration 计算depend 依赖dependence 相关dependent 相关的dependent equations 相关方程组dependent variable 应变数dependent variate 应变量depression 降低depth line 深度线derivability 可微性derivable 可微的derivate 导出数derivation 微分derivative 导数derivative of a distribution 分布导数derivative of a vector 向量导数derivative of higher order 高阶导数derivative of n th order n阶导数derive 导出derived algebra 导出代数derived equation 导出方程derived function 导数derived functor 导函子derived graph 导出图derived rule of inference 推理的导出规则derived series 导出列derived set 推导集derived unit 导出单位derogatory matrix 减次阵descartes rule of signs 笛卡儿正负号规则descending central series 降中心列descending chain 降链descending chain condition 降链条件descending difference 前向差分descending induction 递减归纳descending order 递减次序descending power series 递减幂级数descent 下降descent method 下降法description 描述description operator 摹状算子descriptive form 描述形式descriptive function 描述形式descriptive geometry 画法几何descriptive set theory 描述集论descriptive statistics 描述统计学design 计划design of experiments 实验设计detached coefficients 分离系数determinant 行列式determinant of infinite order 无限行列式determinant of the coefficients 系数行列式determinant of the coefficients of a linear form 线性形式的系数行列式determinantal divisor 行列式因子determinantal equation 行列式方程determinate 一定的determinate automaton 确定性自动机determinate system 确定组determine 决定出determined system 确定组determining equation 决定方程determining factor 决定因素deterministic digital system 确定性数字系统deterministic optimization 确定性最优化deterministic process 确定过程deterministic programming 确定性最优化develop 展开developability 可展性developable 可展的developable function 可展函数developable surface 可展曲面development 展开development in power series 幂级数展开deviate 偏离deviation 偏差deviation from the mean 平均偏差diadic system 二进制数系diagnostic routine 诊断程序diagonal 对角线diagonal continued fraction 对角连分数diagonal dominancy 对角优势diagonal element 对角元素diagonal form 对角型diagonal map 对角映射diagonal matrix 对角阵diagonal method 对角线法diagonal morphism 对角射diagonal of a determinant 行列式的对角线diagonal of the face 面对角线diagonal point 对边点diagonal procedure 对角线法diagonal process 对角线法diagonal sequence 对角序列diagonal sum 矩阵的迹diagonal sum rule 对角求和规则diagonalizable matrix 可对角化矩阵diagonalization 对角线化diagonalize 对角化diagonally dominant matrix 对角占优矩阵diagram 图表diagram scheme 图解概型diameter 直径diameter of a circle 圆的直径diametric plane 径面diamond shaped 菱形的dichotomy 二分法diffeomorphic mapping 微分同胚映射diffeomorphism 微分同胚映射difference 差difference boundary value problem 差分边值问题difference differential equation 差分微分方程difference equation 差分方程difference group 差群difference method 差分法difference operator 差分算子difference product 差积difference quotient 均差difference schema 差分格式difference sequence 差数序列difference set 差集difference table 差分表different 共轭差积differentiability 可微性differentiable 可微的differentiable function 可微函数differentiable manifold of class c c类微分廖differential 微分differential algebra 微分代数differential analyzer 微分分析仪differential and integral calculus 微积分differential calculus 微分学differential circuit 微分电路differential coefficient 微分系数differential cross section 微分截面differential curve 微分曲线differential difference equation 差分微分方程differential equation 微分方程differential equation with delayed argument 延滞方程differential equation with deviating argument 偏差自变数微分方程differential equation with lag 滞后微分方程differential equation with separated variables 分离变数型微分方程differential expression 微分式differential form 微分形式differential form of the first kind 第一种微分形式differential game 微分对策differential geometry 微分几何学differential ideal 微分理想differential method 微分法differential of arc 微弧differential operator 微分算子differential parameter 微分参数differential quotient 微分系数differential ring 微分环differential scattering 微分散射截面differential topology 微分拓扑differentiate 微分differentiating circuit 微分电路differentiation 微分differentiation of a function 函数的微分法differentiation of implicit function 隐函数微分法differentiation operator 微分算子differentiation symbol 微分记号differentiation term by term 逐项微分differentiation theorem 微分定理differentiator 微分器diffraction 衍射diffraction angle 衍射角diffraction curve 衍射曲线diffraction disc 绕射盘diffusion 扩散diffusion coefficient 扩散系数diffusion constant 扩散常数diffusion equation 扩散方程diffusion process 扩散过程digamma function 双函数digit 数字digital 数字的digital computer 数字计算机digital control 数字控制digital differential analyzer 数字微分分析仪digital recorder 数字式自动记录器digital simulation 数据模拟digitize 计数化dihedral angle 二面角dihedral group 二面体群dihedron 二面体dilatation 单项变换dilated maximum principle 扩张极大值原理dilemma 二难推论dimension 量纲dimension theorem 维数定理dimension theory 维数论dimensional 量纲的dimensional analysis 维量分析dimensional equation 量纲方程dimensionality 量纲dimensionless 无量纲的dimensionless quantity 无因次量dimer 二聚物dimetric 二维的diophantine analysis 丢番图分析diophantine equation 丢番图方程diplohedron 扁方二十四面体dirac delta distribution 狄垃克函数dirac equation 狄拉克方程dirac measure 狄拉克测度direct 直接的direct analytic continuation 直接解析开拓direct correspondence 直接对应direct decomposition 直分解direct factor 直积因子direct image 直接象direct limit 归纳极限direct method 直接法direct numerical method 直接数值法direct predecessor 直前仟direct product 直积direct successor 紧接后元direct sum 直和direct system 归纳系direct union 直并directed circuit 有向回路directed distance 有向距离directed edge sequence 有向棱序列directed graph 有向图directed group 有向群directed line 有向元directed line segment 有向线段directed path 有向通路directed quantity 有向量directed set 有向集directed system 有向系directing curve 有向曲线direction 方向direction angle 方向角direction cosine 方向余弦direction field 方向场direction of principal axis 轴方向direction of principal curvature 助率方向direction parameter 方向参数directional 定向的directional derivative 方向导数directional differentiation 方向微分法directional field 方向场directivity 方向性directly proportional 直接比例的directoin search program 方向检颂序director circle 准圆director cone 准锥面director plane 准平面directrix 准线directrix of a conic 二次曲线的准线dirichlet boundary condition 狄利克雷边界条件dirichlet conditions 狄利克雷条件dirichlet distribution 狄利克雷分布dirichlet domain 狄利克雷域dirichlet drawer principle 狄利克雷抽屉原理dirichlet function 狄利克雷函数dirichlet integral 狄利克雷积分dirichlet principle 狄利克雷原理dirichlet problem 狄利克雷问题dirichlet product 狄利克雷乘积dirichlet series 狄利克雷级数dirichlet space 狄利克雷空间dirichlet theorem 狄利克雷定理disagreement 不符合disappearance 消失disassembly 拆卸disc 圆盘disconnected space 不连通空间discontinuity 不连续discontinuity interval 不连续区间discontinuity on the left 左方不连续性discontinuity on the right 右方不连续性discontinuous function 不连续函数discontinuous group 不连续群discontinuous random variable 不连续变量discontinuous set 不连续集discontinuous term 不连续项discontinuous variate 不连续变量discontinuum 密断统discount 折扣discount factor 折扣因子discrete 分立的discrete category 离散范畴discrete continuous system 离散连续系统discrete distribution 离散分布discrete distribution function 离散分布函数discrete flow 离散流discrete fourier transform 离散傅里叶变换discrete group 离散群discrete mathematics 离散数学discrete optimization 离散最佳化discrete optimization problem 离散最优化问题discrete problem 离散问题discrete process 离散随机过程discrete programming 离散规划discrete random variable 离散随机变量discrete series 离散序列discrete set 离散集discrete spectrum 离散谱discrete state 离散状态discrete system 离散系统discrete time 离散时间discrete topological space 离散拓扑空间discrete topology 离散拓扑discrete uniform distribution 离散均匀分布discrete valuation 离散赋值discreteness 离散性discretization 离散化discretization error 离散化误差discrimator 判别式函数discriminant 判别式discriminant analysis 判别分析discriminant function 判别式函数discriminant of a polynomial 多项式的判别式discriminatory analysis 判别分析disjoint elements 不相交元素disjoint relations 不相交关系disjoint sets 不相交集disjoint sum 不相交并集disjoint union 不相交并集disjointed set 不相交集disjunction 析取disjunction sign 析取记号disjunction symbol 析取记号disjunctive normal form 析取范式disjunctive proposition 选言命题disk 圆盘disorder 无秩序disorder order transformation 无序有序变化dispersion 方差dispersion matrix 方差矩阵dispersion relations 分散关系dispersive 扩散的displacement 位移displacement operator 位移算符display statusconcomitant 相伴式disposition 配置disproportion 不相称disproportionate 不成比例的dissection 剖分dissimilar terms 不同类项dissipation 散逸dissipation of energy 消能dissipative function 散逸函数dissipative measurable transformation 散逸可测变换dissipative system 耗散系dissociation 解离dissociation constant 分离常数distance axioms 距离公理distance between two points 两点间距distance circle 距离圆distance function 距离函数distance matrix 距离矩阵distance meter 测距仪distance point 距离点distinction 差别distinguish 辨别distinguished polynomial 特异多项式distortion 畸变distortion angle 畸变角distortion theorem 畸变定理distortionless 无畸变的distributed constant 分布常数distributed parameter 分布参数distribution 分布distribution coefficient 分布系数distribution curve 分布曲线distribution family 分布族distribution function 分布函数distribution law 分布律distribution of prime numbers 素数分布distribution parameter 分布参数distribution ratio 分布系数distribution rule 分布规则distribution space 广义函数空间distribution with negative skewness 负偏斜分布distribution with positive skewness 正偏斜分布distributionfree test 无分布检验distributive 分配的distributive lattice 分配格distributive law 分配律distributivity 分配性disturbance 扰动disturbing function 扰动函数diverge 发散divergence 发散divergence of a series 级数发散divergence of tensor field 张量场的散度divergence of vector field 向量场的散度divergent sequence 发散序列divergent series 发散级数divide 除divided difference 均差dividend 被除数divider compasses 除法器两脚规dividers 除法器两脚规divisibility 可除性divisible 可除的divisible element 可除元素division 除法;划分division algebra 可除代数division algorithm 辗转相除法division of a line segment 线段的分割division ring 可除环division transformation 有剩余的除法division with remainder 有剩余的除法divisor 因divisor class 除子类divisor function 除数函数divisor problem 除数问题documentation 文件编制documentation of program 程序文档dodecagon 十二边形dodecagonal 十二边形的dodecahedral number 十二面体数dodecahedron 十二面体dog curve 追踪曲线domain 定义域domain of attraction 吸引范围domain of convergence 收敛域domain of definition 定义域domain of dependence 依赖域domain of existence 存在域domain of integration 积分区域domain of integrity 整环domain of meromorphy 亚纯域domain of regularity 正则域domain of transitivity 可递域domain of unsolvability 不可解域domain of variability 定义域dominant 帜dominant strategy 优策略dominant weight 最高权dominate 支配dominated convergence 控制收敛dominating set 控制集domination 支配domination principle 优势原理domino problem 多米诺问题dot 点dot chart 点图表dot product 纯量积dotted 点线的dotted line 点线dotted spinor 有点旋量double 双的double angle formulas 倍角公式double chain complex 双链复形double complex 二重复形double cone 对顶锥double coset 重倍集double cusp 双尖点double element 二重元素double exponential distribution 二重指数分布double folium 双叶线double fourier series 二重傅里叶级数double integral 二重积分double laplace transformation 二重拉普拉斯变换double layer 双层double layer potential 双层位势double limit 二重极限double line 二重线double loop 双环路double negation 双重否定double orthogonal system 二重正交系double periodicity 双周期性double plane 二重面double point 重点double point of curve 曲线的二重点double poisson distribution 二重泊松分布double product 二重积double ratio 交比double root 重根double sequence 二重数列double series 二重级数double subscript 双下标double sum 二重和double tangent 二重切线double valued function 双值函数double vector product 二重向量积doubly periodic function 双周期函数dozen 一打draw 拉drum 磁鼓dual abelian variety 对偶阿贝耳簇dual automorphism 逆自同构dual base 对偶基dual basis 对偶基dual category 对偶范畴dual cell 对偶胞腔dual complex 对偶复形dual cone 对偶锥dual curve 对偶曲线dual figure 对偶图dual form 对偶形式dual formula 对偶公式dual graph 对偶图dual group 特贞群dual ideal 对偶理想dual isomorphism 对偶同构dual lattice 对偶格dual mapping 对偶映射dual module 对偶模dual number 对偶数dual operation 对偶运算dual operator 对偶算子dual problem 对偶问题dual relation 对偶关系dual representation 对偶表示dual simplex method 对偶单形法dual spaces 对偶空间dual system 对偶系统dual theorem 对偶定理dual vector space 对偶向量空间duality 对偶性duality principle 对偶原理duality relation 对偶关系duality theorem 对偶定理duel 竞赛dummy index 哑指标duodecimal notation 十二进记数法duodecimal system 十二进制duodecimal system of numbers 十二进数系duplication formula 倍角公式duplication of the cube 倍立方duration 持久时间dyad 并向量dyadic expansion 二进展开dyadic product 并向量积dyadic rational 二进有理数dynamic optimization 动态最优化dynamic programming 动态规划dynamic store 动态存储器dynamic system 动力系统dynamical variables 动态变数dynamics 力学dynkin diagram 丹金图形。
- 1、下载文档前请自行甄别文档内容的完整性,平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益,请联系客服反馈,我们会尽快为您处理(人工客服工作时间:9:00-18:30)。
Learning Dense3D CorrespondenceFlorian Steinke∗,Bernhard Sch¨o lkopf∗,Volker Blanz+∗Max Planck Institute for Biological Cybernetics,72076T¨u bingen,Germany{steinke,bs}@tuebingen.mpg.de+Universit¨a t Siegen,57068Siegen,Germanyblanz@mpi-sb.mpg.deAbstractEstablishing correspondence between distinct objects is an important and non-trivial task:correctness of the correspondence hinges on properties which aredifficult to capture in an a priori criterion.While previous work has used a prioricriteria which in some cases led to very good results,the present paper exploreswhether it is possible to learn a combination of features that,for a given trainingset of aligned human heads,characterizes the notion of correct correspondence.By optimizing this criterion,we are then able to compute correspondence andmorphs for novel heads.1IntroductionEstablishing3D correspondence between surfaces such as human faces is a crucial element of class-specific representations of objects in computer vision and graphics.On faces,for example,cor-responding points may be the tips of the noses in3D scans of different individuals.Dense corre-spondence is a mapping or”warp”from all points of a surface onto another surface(in some cases, including the present work,extending from the surface to the embedding space).Once this mapping is established,it is straightforward,for instance,to compute morphs between objects.More im-portantly,if correspondence mappings between a class of objects and a reference object have been established,we can represent each object by its mapping,leading to a linear representation that is able to describe also new objects of similar shape and texture(for further details,see[1]).The practical relevance of surface correspondence has been increasing over the last years.In com-puter graphics,applications involve morphing,shape modeling and animation.In computer vision, an increasing number of algorithms for face and object recognition based on2D images or3D scans, as well as shape retrieval in databases and3D surface reconstruction from images,rely on shape rep-resentations that are built upon dense surface correspondence.Unlike existing algorithms that define some ad-hoc criteria for identifying corresponding points on two objects,we treat correspondence as a machine learning problem and propose a data-driven approach that learns the relevant criteria from a dataset of given object correspondences.In stereo vision and opticalflow[2,3],a correspondence is correct if and only if it maps a point in one scene to a point in another scene which stems from the same physical point.In contrast,corre-spondence between different objects is not a well-defined problem.When two faces are compared, only some anatomically unique features such as the corners of the eyes are clearly corresponding, while it may be difficult to define how smooth regions,such as the cheeks and the forehead,are supposed to be mapped onto each other.On a more fundamental level,however,even the problem of matching the eyes is difficult to cast in a formal way,and in fact this matching involves many of the basic problems of computer vision and feature detection.In a given application,the desired correspondence can be dependent on anatomical facts,measures of shape similarity,or the overall layout of features on the surface.However,it may also depend on the properties of human percep-tion,on functional or semantic issues,on the context within a given object class or even on socialconvention.Due to the problematic and challenging nature of the correspondence problem,our cor-respondence learning algorithm may be a more appropriate approach than existing techniques,as it is often easier to provide a set of examples of the desired correspondences than a formal criterion for correct correspondence.In a nutshell,the main idea of our approach is as follows.Given two objects O1and O2,we are seeking a correspondence mappingτsuch that certain properties of x(relative to O1)are preserved inτ(x)(relative to O2)—they are invariant.These properties depend on the object class and as explained above,we cannot hope to characterize them comprehensively a priori.However,if we are given examples of correct and incorrect correspondences,we can attempt to learn properties which are invariant for correct correspondences,while for incorrect correspondences,they are not.We shall do this by providing a dictionary of potential properties(such as geometric features,or texture properties)and approximating a“true”property characterizing correspondence as an expansion in that dictionary.We will call this property warp-invariant feature and show that its computation can be cast as a problem of oriented PCA.The remainder of the paper is structured as follows:in Section2we review some related work, whereas in Section3we set up our general framework for computing correspondencefields.Fol-lowing this,we explain in Section4how to learn the characteristic properties for correspondence and continue to explain two new feature functions in Section5.We give implementation details and experimental results in Section6and conclude in Section7.2Related WorkThe problem of establishing dense correspondence has been addressed in the domain of2D images, on surfaces embedded in3D space,and on volumetric data.In the image domain,correspondence from opticalflow[2,3]has been used to describe the transformations of faces with pose changes and facial expressions[4],and to describe the differences in the shapes of individual faces[5].An algorithm for computing correspondence on parameterized3D surfaces has been introduced for creating a class-specific representation of human faces[1]and bodies[6].[7]propose a method that is designed to align three dimensional medical images using a mutual information criterion. Another interesting approach is[8]:they formulate the problem in a probabilistic setup and then apply standard graphical model inference algorithms to compute the correspondence.Their mesh based method uses a smoothness functional and features based on spin images.See the review[9] for an overview of a wide range of additional correspondence algorithms.Algorithms that are applied to3D faces typically rely on surface parameterizations,such as cylin-drical coordinates,and then compute opticalflow on the texture map as well as the depth image [1].This algorithm yields plausible results,to which we will compare our method.However,the approach cannot be applied unless a parameterization is possible and the distortions are low on all elements of the object class.Even for faces this is a problem,for example around the ears,which makes a more general real3D approach preferable.One such algorithm is presented in[10]:here, the surfaces are embedded into the surrounding space and a3D volume deformation is computed. The use of the signed distance function as a guiding feature ensures correct surface to surface map-pings.We build on this approach that is more closely presented in Section3.A common local geometric feature is surface curvature.Though implicit surface representations allow the extraction of such features[11],these differential geometric properties are inherently in-stable with respect to noise.[12]propose a related3D geometric feature based on integrals and thus more stable to compute.We present a slightly modified version thereof which allows for a much easier computation of this feature from a signed distance function represented as a kernel expansion in comparison to a complete space voxelisation step required in[12].3General Framework For Computing CorrespondenceIn order to formalize our understanding of correspondence,let us assume that all the objects O of class O are embedded in X⊆R3.Given a reference object O r and a target O t the goal of computing a correspondence can then be expressed as determining the deformation functionτ:X→X which maps each point x∈X on O r to its corresponding pointτ(x)on O t.We further assume that we can construct a dictionary of so-called feature functions f i :X →R ,i =1,..,n capturing certain characteristic properties of the objects.[10]propose to use the signed distance function,which maps to each point x ∈X the distance to the objects surface —with positive sign outside the shape and negative sign inside.They also use the first derivative of the signed distance function,which can be interpreted as the surface normal.In section Section 5we will propose two additional features which are characteristic for 3D shapes,namely a curvature related feature and surface texture.We assume that the warp-invariant feature can be represented or at least approximated by an expan-sion in this dictionary.Let γ:X →R n be a weighting function describing the relative importance of the different elements of the dictionary at a given location in X .We then express the warp-invariant feature as f γ:X →R ,f γ(x )= n i =1γi (x )f i (x )with feature functions f i that are object specific;for the target object there is a slight modification in that the space-variant weighting γ(x )needs to refer to the coordinates of the reference object if we want to avoid comparing apples and oranges.We thus use f t γ(x )= n i =1γi (τ−1(x ))f t i (x ),where we never have to evaluate τ−1since we will only require f t γ(τ(x ))below.To determine a mapping τwhich will establish correct correspondences between x and τ(x ),we minimize the functional C reg τ 2H +Xf r γ(x )−f t γ(τ(x )) 2dµ(x )(1)The first term expresses a prior belief in a smooth deformation.This is important in regions where the objects are not sufficiently characteristic to specify a good correspondence.As we will use a Support Vector framework to represent τ,smoothness can readily be expressed as the RKHS norm τ H of the non-parametric part of the deformation function τ(see Section 6).The second term measures the local similarity of the warp-invariant feature function extracted on the reference object f r and on the target object f t and integrates it over the volume of interest.This formulation is a modification of [10]where two feature functions were chosen a priori (the signed distance and its derivative)and used instead of f γ.The motivation for this is that for a correct morph,these functions should be reasonably invariant.In contrast,the present approach starts from the notion of invariance and estimates a location-dependent linear combination of feature functions with a maximal degree of invariance for correct correspondences (cf.next section).We consider location-dependent linear combinations since one cannot expect that all the feature functions that define correspondence are equally important for all points of an object.For example color may be more characteristic around the lips or the eyes than on the forehead.This comes at the cost,however,of increasing the number of free parameters,leading to potential difficulties when performing model selection.As discussed above,it is unclear how to characterize and evaluate correspondence in a principled way.The authors of [10]propose a strategy based on a two-way morph:they first compute a deformation from the reference object to the target,and af-terwards vice versa.A necessary condition for a correct morph is then that the concatenation of the two deformations yield a mapping close to the identity.1Although this method can provide a partial quality criterion even when no ground truth is available,all model selection approaches based on such a criterion need to minimize (1)many times and the computation of a gradient with respect to the parameters is usually not possible.As the minimization is typically non-convex and rather expensive,the number of free parameters that can be optimized is small.For locally varying param-eters as proposed here such an approach is not practical.We thus propose to learn the parameters from examples using an invariance criterion proposed in the next section.4Learning the optimal feature functionWe assume that a database of D objects that are already in correspondence is available.This could for example be achieved by manually picking many corresponding point pairs and training a regres-sion to map all the points onto each other,or by (semi-)automatic methods optimized for the given object class (e.g.,[1]).We can then determine the optimal approximation of the warp-invariant1It is not a sufficient condition,since the concatenation of,say,two identity mappings will also yield the identity.feature function (as defined in the introduction)that characterizes correspondence using the basic features in our dictionary.The warp-invariant feature function should be such that it varies little or not at all for corresponding points,but its value should not be preserved (and have large variance)for random non-matching points.To approximate it,we propose to maximize the ratio of these variances over all weighting functions γ.Thus for each point x ∈X ,we maximizeE d,z d f r γ(x )−f d γ(z d ) 2E d f r γ(x )−f d γ(τd (x ))2(2)Here,f r γ,f d γare the warp-invariant feature functions evaluated on the reference object and the d -th database object respectively.τd (x )is the point matching x on the d -th database object and z d is a random point sampled from it.We take the expectations over all objects in our database,as well as non corresponding points randomly sampled from the objects.Because of the linear dependence of f γon γone can rewrite the problem as the maximization ofγ(x )T C z (x )γ(x )γ(x )T C τ(x )γ(x )(3)with the empirical covariances[C τ(x )]i,j=D d =1 f r i (x )−f d i (τd (x ) f r j (x )−f d j (τd (x )) T ,(4)[C z (x )]i,j =D d =1N k =1f r i (x )−f d i (z d,k ) f r j (x )−f d j (z d,k ) T ,(5)where we have drawn N random sample points from each object in the database.This problem is known as oriented PCA [13],and the maximizing vector γ(x )can be determined by solving the generalized eigenvalue problem C τ(x )v (x )=λ(x )C z (x )v (x ).If v (x )is the normalized eigenvector corresponding to the maximal eigenvalue λ(x ),we obtain the optimal weight vector γ(x )=˜λ(x )v (x )using the scale factor ˜λ(x )= v (x )T C τ(x )v (x ) −1/2.Note that by using this scale factor ˜λ(x ),the contribution of the feature function f γin the objective (1)will vary locally compared to the regularizer:as τ(x )is somewhat arbitrary during the optimiza-tion of (1)the average local contribution will then approximately equal E d,z d f r γ(x )−f d γ(z d ) 2=λ(x ).This implies that if locally there exists a characteristic combination of features —λ(x )is high —it will have a big influence in (1).If not,the smoothness term τ H gets relatively more weight implying that the local correspondence is mostly determined through more global contributions.Note,moreover that while we have described the above for the leading eigenvector only,nothing prevents us from computing several eigenvectors and stacking up the resulting warp-invariant feature functions f 1γ,f 2γ,...,f m γinto a vector valued warp-invariant feature function f γ:X →R m which then is plugged into the optimization problem (1)using the two norm to measure deviations instead of the squared distance.5Basic Feature FunctionsIn our dictionary of basic feature functions we included the signed distance function and its deriva-tive.We added a curvature related feature,the ”signed balls”,and surface texture intensity.5.1Signed BallsImagine a point x on a flat piece of a surface.Take a ball B R (x )with radius R centered at that point and compute the average of the signed distance function s :X →R over the ball’s volume:I s (x )=1V B R (x ) B R (x )s (x )dx −s (x )(6)If the surface around x is flat on the scale of the ball,we obtain zero.At points where the surface is bent outwards this value is positive,at concave points it is negative.The normalization to the valueB 4mm B 28mmC 0mm C 5mm C 15mmFigure 1:The two figures on the left show the color-coded values of the ”signed balls”feature at different radii R .Depending on R ,the feature is sensitive to small-scale structures or large-scale structures only.Convex parts of the surface are assigned positive values (blue),concave parts negative (red).The three figures on the right show how the surface feature function that was trained with texture intensity extends off the surface (for clarity visualized in false colors)and becomes smoother.In the figure,the function is mapped on surfaces that are offset by 0,5and 15mm.of the signed distance function at the center of the ball allows us to compute this feature function also for off-surface points,where the interpretation with respect to the other iso-surfaces does not change.Due to the integration,this feature is stable with respect to surface noise,while mean curvature in differential geometry may be affected significantly.Moreover,the integration involves a scale of the feature.We propose to represent the implicit surface function as in [10]where a compactly supported kernel expansion is trained to approximate the signed distance.In this case the integral and the kernel summation can be interchanged,so we only need to evaluate terms of the form B R (x )k (x i ,x )dx and then add them in the same way as the signed distance function is computed.The value of this basic integral only depends on the distance between the kernel center x i and the test point x .It is compactly supported if the kernel k is.Therefore,we propose to pre-compute these values numerically for different distances and store them in a small lookup table.For the final expansion summation we can then just interpolate the two closest values.We obtained good interpolation results with about ten to twenty distance values.For the case where the surface looks locally like a sphere it is easy to show that in the limit of small balls the value of the ”signed balls”feature function is related to the differential geometric mean curvature H by I s (x )=3π20H 2R 2+O (R 3).5.2Surface properties —TextureThe volume deformation approach presented in Section 3requires the use of feature functions de-fined on the whole domain X .In order to include information f |∂Ωwhich is just given on a surface ∂Ωof the object whose interior volume is Ω,e.g.the texture intensity,we propose to extended the surface feature f |∂Ωinto a differentiable feature function f :X →R such that f →f |∂Ωas we get closer to the surface.At larger distances from the surface,f should be smoother and tend towards the mean feature value.This is a desirable property during the optimization of (1)as it helps to avoid local minima.Finally,the feature function f and its gradient should be efficient to evaluate.We propose to use a multi-scale compactly supported kernel regression to determine f :at each scale,from coarse to fine,we select approximately equally spaced points on the surface at a distance related to the kernel width of that scale.Then we compute the feature value at these points averaged over a sphere of radius of the corresponding kernel support.With standard quadratic SVR regression we fit the remainder of what was achieved on larger scales to the training values.Due to the sub-sampling the kernel regressions do not contain too many kernel centers and the compact support of the kernel ensures sparse kernel matrices.Thus,efficient regression and evaluation is guaranteed.Because all kernel centers lie on the surface and reach to different extents into the volume X depending on the kernel size of their scale,we can model small-scale variations on the surface and close to it,whereas the regression function varies only on a larger scale further away from the surface.6ExperimentsImplementation.In order to optimize (1)we followed the approach of [10]:we represent the deformation τas a multi-scale compactly supported kernel expansion,i.e.,the j -th component,empty C B N horiz.N vert.N depth.Figure 2:Locations that are marked yellow show an above threshold,relative contribution (see text)of a given feature in the warp-invariant feature function.C is the surface intensity feature,B the signed balls feature (R =6mm ),N the surface normals in different directions.Note that points where color has a large contribution (yellow points in C)are clustered around regions with characteristic color information,such as the eyes or the mouth.j =1,2,3,of τis τj (x )=x j + S s =1 N s i =1αj i,s k (x,x i,s )with the compactly supported kernel function k :X ×X →R .The regularizer then is τ 2H := S s =1 3j =1 N si,l =1αj i,s αj l,s k (x l,s ,x i,s ).We approximate the integral in (1)by sampling N s kernel centers x i,s on each scale s =1,...,Saccording to the measure µ(x )and minimize the resulting non-linear optimization problem in the coefficients αj i,s for each scale from coarse to fine using a second order Newton-like method [14].As a test object class we used 3D heads with known correspondence [1].100heads were used for the training object database and 10to test our correspondence algorithm.As a reference head we used the mean head of the database.The faces are all in correspondence,so we can just linearly average the vertex positions and the texture images.However,the correspondence of the objects in the database is only defined on the surface.In order to extend it to the off-surface points x i,s ,we generated these locations by first sampling points from the surface and then displacing them along their surface normals.This implied that we were able to identify the corresponding points also on other heads.For each kernel center x i,s ,we learned the weighting vector γ(x i,s )as described in Section 4.In one run through the database we computed for each head the values of all proposed basic feature functions for all locations,corresponding to kernel centers on the reference head,as well as for 100randomly sampled points z .The points z should be typical for possible target locations τ(x i,s )during the optimization of (1).Thus,we sampled points up to distances to the surface proportional to the kernel widths used for the deformation τ.We then estimated the empirical covariance matrices for each kernel center yielding the weight vectors via a small eigenvalue decomposition of size n ×n where n is the number of used basic features.The parameters C reg —one for each scale —were determined by optimizing computed deformation fields from the reference head to some of the training database heads.We minimized the mismatch to the correspondence given in the database.Feature functions.In Figure 1,our new feature functions are visualized on an example head.Each feature extracts specific plausible information,and the surface color can be extended off the surface.Learned weights.In Figure 2,we have marked those points on the surface where a given feature has a high relative contribution in the warp-invariant feature function.As a measure of contribution we took the component of the weight vector γ(x i,s )that corresponds to the feature of interest and multiplied it with the standard deviation of this feature over all heads and all positions.Note that the weight vector is not invariant to rescaling the basic feature functions,unlike the proposed measure.Finally,we normalized the contributions of all features at a given point x i,s to sum to one,yielding the relative contribution.In the table below the relative contribution of each feature is listed.S N horiz.N vert.N depth.C B 8%B 3%average rel.contribution0.8320.0920.0230.0380.0080.0060.003max rel.contribution 0.9970.7010.4290.4460.3940.2720.333Here and below,S is signed distance,N surface normals,C the proposed surface feature function trained with the intensity values on the faces,and B is the ”signed balls”feature with radii given by the percentage numbers scaled to the diameter of the head.The signed distance function is the best preserved feature (e.g.all surface points take the value zero up to small approximation errors).The resulting large weight of this feature is plausible as a surface-to-surface mapping is a necessary condition for a morph.However,combined with Figure 2Reference Deformed Target Deformed TargetFigure3:The average head of the database–the reference–is deformed to match four of the target heads of the test set.Correct correspondence deforms the shape of the reference head to the target face with the texture of the mean face well aligned to the shape details.we can see that the method assigns plausible non-zero values also to other features where these can be assumed to be most characteristic for a good correspondence.Correspondence.We applied our correspondence algorithm to compute the correspondence to the test set of10heads.Some example deformations are shown in Figure3for a dictionary consisting of S,N(hor,vert,depth)C,B(radii3%and8%).Numerical evaluation of the morphs is difficult. We compare our method with the results of the correspondence algorithm of[1]on points that are uniformly drawn form the surface(first column)and for24selected marker points(second column). These markers were placed at locations around the eyes or the mouth where correspondence can be assumed to be better defined than for example on the forehead.Still,the error made by humans when picking these positions has turned out to be around1–2mm.The table below shows mean results in mm for different settings.uniform markers error signed distance(a)all weights equal 5.97 4.49 1.49(b)our method(independent of x) 3.74 1.480.05(c)our method(1eigenvector) 3.74 1.340.04(d)our method(2eigenvectors) 3.62 1.190.04(e)our method(4eigenvectors) 3.56 1.110.04(f)our method(6eigenvectors) 3.55 1.100.04(g)our method(1eigenvector,without B,C) 3.76 1.420.04If all weights are equal independent of location or feature(a),the result is not acceptable.A care-ful weighting of each feature separately,but independent of location(b)—as could potentially be achieved by[10]—improves the quality of the correspondence.To obtain these weights we av-eraged the covariance matrices C z(x),Cτ(x)over all points and applied the proposed algorithm in Section4,but independent of x.However,a locally adapted weighting(c)outperforms the above methods and using more than one eigenvector(d-f)further enhances the correspondence.Note that although the results are not identical to[1],our algorithm’s accuracy is consistent with the human labeling on the scale of the latter’s presumed accuracy(1-2mm).For uniformly sampled points,the differences are slightly larger(4mm),but we need to bear in mind that that algorithm’s results cannot be considered ground truth.Experiment(g)which is identical to(c)but with the color and signed balls feature omitted demonstrates the usefulness of these additional basic feature functions. Computation times ranged between5min and one hour and depended significantly on the number of scales used(here4),the number of kernel centers generated,and the number of basic features included in the dictionary.For large radii R the signed balls feature becomes quite expensive to compute,since many summands of the signed distance function expansion have to be accumulated. Our method to select the important features for each point in advance,i.e.before the optimization isReference 25%50%75%TargetFigure 4:A morph between a human head and the head of the character Gollum (available from ).As Gollum’s head falls out of our object class (human heads),we assisted the training procedure with 28manually placed markers.started,would allow for a potentially high speed-up:At locations where a certain feature has a very low weight,we could just omit it in the evaluation of the cost function (1).7ConclusionWe have proposed a new approach to the challenging problem of defining criteria that characterize a valid correspondence between 3D objects of a given class.Our method learns an appropriate criterion from examples of correct correspondences.The approach thus applies machine learning to computer graphics at the early level of feature construction.The learning technique has been implemented efficiently in a correspondence algorithm for textured surfaces.In the future,we plan to test our method with other object classes.Even though we have concentrated in our experiments on 3D surface data,the method may be applicable also in other fields such as to align CT or MR scans in medical imaging.It would also be intriguing to explore the question whether our paradigm of learning the features characterizing correspondences might reflect some of the cognitive processes that are involved when humans learn about similarities within object classes.References[1]V .Blanz and T.Vetter.A morphable model for the synthesis of 3d faces.In SIGGRAPH’99ConferenceProceedings ,pages 187–194,Los Angeles,1999.ACM Press.[2] B.D.Lucas and T.Kanade.An iterative image registration technique with an application to stereo vision.In IJCAI81,pages 674–679,1981.[3] B.K.P.Horn and B.G.Schunck.Determining optical flow.Artif.Intell.,17(1-3):185–203,1981.[4] D.Beymer and T.Poggio.Image representations for visual learning.Science ,272:1905–1909,1996.[5]T.Vetter and T.Poggio.Linear object classes and image synthesis from a single example image.IEEETrans.on Pattern Analysis and Machine Intelligence ,19(7):733–742,1997.[6] B.Allen,B.Curless,and Z.Popovic.The space of human body shapes:reconstruction and parameteri-zation from range scans.In Proc.SIGGRAPH ,pages 612–619,2002.[7] D.Rueckert and A.F.Frangi.Automatic construction of 3-d statistical deformation models of the brainusing nonrigid registration.IEEE Trans.on Medical Imaging ,22(8):1014–1025,2003.[8] D.Anguelov,P.Srinivasan,H.-C.Pang,D.Koller,S.Thrun,and J.Davis.The correlated correspondencealgorithm for unsupervised registration of nonrigid surfaces.In Neural Information Processing Systems 17,pages 33–40.MIT Press,2005.[9]M.Alexa.Recent advances in mesh puter Graphics Forum ,21(2):173–196,2002.[10] B.Sch¨o lkopf,F.Steinke,and V .Blanz.Object correspondence as a machine learning problem.InProceedings of the 22nd International Conference on Machine Learning (ICML 05),July 2005.[11]J.-P.Thirion and A puting the differential characteristics of isointensity surfaces.Journalof Computer Vision and Image Understanding ,61(2):190–202,March 1995.[12]N.Gelfand,N.J.Mitra,L.J.Guibas,and H.Pottmann.Robust global registration.In Proc.EurographicsSymposium on Geometry Processing ,pages 197–206,2005.[13]K.I.Diamantaras and S.Y .Kung.Principal component neural networks:theory and applications .JohnWiley &Sons,Inc.,1996.[14] D.C.Liu and J.Nocedal.On the limited memory bfgs method for large scale optimization.Math.Program.,45(3):503–528,1989.。