Sei kuzivikanwa kwemufananidzo kwakakosha?

Inenge 80 muzana yezvirimo painternet zvinoonekwa. Iwe unogona kutotanga kushanda kuti nei tagging yemifananidzo ingabata nzvimbo yayo samambo wetafura yemukati. Kunyangwe vari vanhu kana makambani, kucherechedzwa kwemufananidzo weAI kwaita kuti zvikwanise kuona zvinoonekwa pamhepo nekupokana kushoma. Ikoko kwakapoteredza 657 bhiriyoni mapikicha anotumirwa gore rega rega digitally, nehuwandu huchionekwa pasocial media. Chikamu chakanaka chemifananidzo iyi vanhu vari kusimudzira zvigadzirwa, kunyangwe ivo vari kuzviita vasingazivi. Mushandisi-yakagadzirwa zvemukati (UGC) mune yayo yakachena fomu inogonesa yakanakisa mabhureki sezvo ichipa yakanakisa mhando yekusimudzira.
Kune maturusi ekushambadzira ekuyambira makambani kana paine kutaurwa kwevatengi pasocial media, asi ko kana kukwidziridzwa kwema brand kunoitika pasina chero munhu anomaka zita rake musocial post? Apa ndipo apo AI kucherechedzwa kwemufananidzo kunoratidza kukosha kwayo. Kana iyo tekinoroji yakadyiswa iwo chaiwo dhataseti, AI inogona kuona mufananidzo usina chaiwo tag inotaurwa. Mhedzisiro yacho yakakosha kune ma brand kuti ateedzere uye ateedzere mataurirwo aanoitwa munharaunda.

Kuzivikanwa kwemufananidzo kunoshanda sei?

Sezvatinoziva AI inogona kutsvaga social media mapuratifomu ichitsvaga mafoto uye nekuaenzanisa neakawanda data seti. Inobva yasarudza pamufananidzo wakakodzera unofanana pamwero nekukurumidza zvakanyanya kupfuura izvo vanhu vanogona. MaBrand anoshandisa kucherechedzwa kwemifananidzo kuti awane zvirimo zvakafanana nezvavo pasocial media. Izvi zvinoreva kuzivisa chiratidzo chemhando kana kuziva kuiswa kwechigadzirwa kwakaiswa pakati pevashandisi vesocial media. Kukumbira vanhu kuti vaongorore ruzivo rwakawanda zviri nyore kunonetesa. AI hainetseki nekukanganisa kwevanhu, uye inodzosa mibairo chaiyo pamazinga asingaenzaniswi. AI yekuzivikanwa kwemufananidzo inotarisisa zviri kutaurwa nevanhu nezve brand pasina kudiwa kwemavara. Brands inokwanisa kuteedzera mataurirwo avo emagariro pasina vashandisi vanoda kutaipa zita rekambani vanozozviwana vari munzvimbo ine mukana. Iko kugona kupinza muvhavha yavo yepamhepo chete kuburikidza neAI inozivikanwa zviziviso yakakura uye inopa isingaenzaniswi yekuvhara.

Heano mamwe akajairika mabasa ekuzivikanwa kwemufananidzo: -

Pakutanga isu tinofanirwa kuona kana iyo data yemufananidzo ine chimwe chinhu chakasarudzika, chimiro, kana chiitiko. Iri basa rinowanzogadziriswa zvakasimba uye pasina kushanda nesimba nemunhu, asi harisati ragadziriswa zvinogutsa muchiratidzo chekombuta yenyaya yakajairwa: zvinhu zvinopokana mumamiriro ezvinhu. Nzira dziripo dzekugadzirisa dambudziko iri dzinogona kugadziriswa zvakanyanya chete pazvinhu zvakanangana, senge zvinhu zviri nyore zvejometri (semuenzaniso, polyhedra), zviso zvevanhu, mavara akadhindwa kana akanyorwa nemaoko, kana mota, uye mune mamwe mamiriro, anowanzo kutsanangurwa maererano nemashoko. yekujekesa kwakanyatsotsanangurwa, kumashure, uye chimiro chechinhu chinoenderana nekamera. Mhando dzakasiyana dzedambudziko rekuzivikanwa dzinotsanangurwa mumabhuku:

• Kuzivikanwa kwechinhu

Chinhu chimwe kana chinoverengeka chakafanotsanangurwa kana chakadzidzwa zvinhu kana makirasi echinhu chinogona kuzivikanwa, kazhinji pamwe chete nenzvimbo dzazvo dze2D mumufananidzo kana 3D inomira panzvimbo.

• Kuzivikanwa
Chiitiko chega chega chechinhu chinozivikanwa. Mienzaniso kutariswa kwechiso chemumwe munhu kana zvigunwe, kana kuzivikanwa kwemotokari chaiyo.

• Kuonekwa
Iyo data yemufananidzo inotariswa kune chaiyo mamiriro. Mienzaniso ndeyekuonekwa kwezvingangoitika maseru kana matishu mumifananidzo yezvokurapa kana kuonekwa kwemotokari muotomatiki mugwagwa wetoll system. Kuona kwakavakirwa pazviri nyore uye nekukurumidza computations dzimwe nguva kunoshandiswa kutsvaga matunhu madiki einonakidza data data iyo inogona kuongororwa zvakare nehunyanzvi hunoda hunyanzvi hwekuita dudziro chaiyo.

Mabasa akati wandei anoenderana nekuzivikanwa aripo, akadai se:

• Content-based image kudzorerwa
Pano kuwana mifananidzo yese mune yakakura seti yemifananidzo ine chaiyo yemukati. Zvirimo zvinogona kutsanangurwa nenzira dzakasiyana, semuenzaniso maererano nekufanana kwemufananidzo waunovavarira (ndipe mifananidzo yese yakafanana nemufananidzo X), kana maererano nedanho repamusoro rekutsvaga rinopihwa sekuisa mavara (ndipe mifananidzo yese ine dzimba dzakawanda, dzinotorwa munguva yechando, uye hadzina motokari mairi).

• Pose fungidziro
isu tinofanirwa kufungidzira chinzvimbo kana kutariswa kwechinhu chakanangana nekamera. Muenzaniso wekushandisa kwehunyanzvi uhu kunenge kuri kubatsira robhoti kudzoreredza zvinhu kubva mubhandi rekutakura munzvimbo yegungano.

• Kuzivikanwa kwechimiro chemaziso
OCR iyo iri kuzivisa mavara mumifananidzo yeakadhindwa kana nemaoko akanyorwa, kazhinji nechinangwa chekuisa kodhi zvinyorwa mufomati zvakanyanya uye kugonesa kugadzirisa kana kuisa indexing Dhipatimendi reComputer Science neEngineering, Michigan State University. "Pateni Recognition uye Image Processing (PRIP) Lab fakaroti uye vadzidzi vanoongorora kushandiswa kwemichina kuziva mapatani kana zvinhu. Nzira dzakagadzirwa kuti dzinzwe zvinhu, kuona kuti ndezvipi zvezvimiro zvazvo zvinozvisiyanisa kubva kune vamwe, uye kugadzira algorithms inogona kushandiswa nemuchina kuita kupatsanura. Zvishandiso zvakakosha zvinosanganisira kuzivikanwa kwechiso, kuzivikanwa kwezvigunwe, kuongororwa kwemufananidzo wegwaro, 3D chinhu chekuvaka modhi, marobhoti navigation, uye kuona / kuongorora 3D volumetric data. Matambudziko azvino ekutsvagisa anosanganisira biometric authentication, otomatiki kuongorora uye kuronda, isina maoko HCI, kuenzanisira kumeso, digital watermarking uye kuongorora chimiro chemagwaro epamhepo. Vachangopedza kudzidza murabhoritari vakashanda mukuzivikanwa kwekunyora nemaoko, kusimbisa siginecha, kudzidza kwekuona, uye kutora mifananidzo. "

⦁ Kuzivikanwa Kwechiso
tinoziva kuti chiso recognition masisitimu ari kuramba achizivikanwa senzira dzekubvisa ruzivo rwebiometric. Kuzivikanwa kwechiso kune basa rakakosha mune biometric masisitimu uye inokwezva kune akawanda maapplication anosanganisira yekuona yekutarisa uye chengetedzo. Nekuda kwekugamuchirwa kweruzhinji kwemifananidzo yechiso pamagwaro akasiyana, kuzivikanwa kwechiso kune mukana mukuru wekuva chizvarwa chinotevera chebiometric tekinoroji yesarudzo.

Image Recognition Systems

⦁ Kuongorora kwekufamba
Mabasa akati wandei ane chekuita nefungidziro yekufamba uko kutevedzana kwemufananidzo kunogadziriswa kuti kuburitse fungidziro yekumhanya kungave panzvimbo yega yega pamufananidzo kana muchiitiko che3D, kana kunyangwe kamera inogadzira mifananidzo. Mienzaniso yemabasa akadai ndeiyi:

⦁  Ego motion
Kuona kuomarara kwe 3D (kutenderera neshanduro) kwekamera kubva mukutevedzana kwemufananidzo unogadzirwa nekamera.

⦁ Kutsvaga
Kutevera kunotevera mafambiro e (kazhinji) diki seti yemapoinzi kana zvinhu (semuenzaniso, mota kana vanhu) mukutevedzana kwemufananidzo.

⦁ Kuyerera kwemaziso
Izvi ndezvekuona, pane imwe neimwe yemufananidzo, kuti iyo nzvimbo iri kufamba sei maererano nendege yemufananidzo, kureva, kufamba kwayo kunooneka. Kufamba uku mhedzisiro yekufamba kweiyo 3D point iri kufamba munzvimbo uye kuti kamera iri kufamba sei inoenderana nechiitiko.

⦁ Kuvakazve nzvimbo
Kupihwa imwe kana (kazhinji) mimwe mifananidzo yechiitiko, kana vhidhiyo, kuvakwazve kwechiitiko kune chinangwa chekombuta ye 3D modhi yechiitiko. Muchiitiko chakareruka iyo modhi inogona kuve seti ye3D mapoinzi. Dzimwe nzira dzakaomarara dzinoburitsa yakazara 3D yepamusoro modhi

⦁ Kudzoreredza mufananidzo
Chinangwa chekudzorera mufananidzo kubvisa ruzha (sensor ruzha, blur yekufamba, nezvimwewo) kubva pamifananidzo. Iyo yakapfava nzira yekubvisa ruzha mhando dzakasiyana dzemasefa senge yakaderera-pass mafirita kana epakati mafirita. Dzimwe nzira dzakaoma kunzwisisa dzinotora modhi yemagadzirirwo emifananidzo yenzvimbo inotaridzika, modhi inovasiyanisa kubva kune ruzha. Nekutanga kuongorora iyo data yemufananidzo maererano nemagadzirirwo emifananidzo yenzvimbo, semitsetse kana micheto, uyezve kudzora kusefa kunoenderana neruzivo rwenzvimbo kubva padanho rekuongorora, nhanho iri nani yekubvisa ruzha inowanzowanikwa kana ichienzaniswa nemaitiro akareruka. Muenzaniso mundima iyi mufananidzo wavo. Mamwe masisitimu anomira-ega maapplication anogadzirisa imwe kuyerwa kana dambudziko rekuona, nepo mamwe achiumba sub-system yedhizaini yakakura iyo, semuenzaniso, inewo ma sub-system ekudzora ma mechanic actuators, kuronga, ruzivo dhatabhesi, munhu- Michina yekubatanidza, nezvimwewo. Kuitwa chaiko kwekombiyuta yekuona system kunoenderana nekuti kushanda kwayo kwakafanotsanangurwa here kana kuti chimwe chikamu chayo chinogona kudzidzwa kana kugadziridzwa panguva yekushanda. Kune, zvisinei, maitiro akajairika anowanikwa mune akawanda macomputer ekuona masisitimu.

 

Kudzidza kwakadzama nekuzivikanwa kwemufananidzo

Kuzivikanwa kwemufananidzo kwaive kwakapoteredza AI isati yasvika. Zvakadaro muchina kudzidza chinhu kushandura nzira dzekuziva chinhu kana chiso chemunhu. Kudzidza kwemuchina kunoshanda chete kana paine data yekuidyisa, zvisinei. Kune ese otomatiki eAI, kuita basa rekuona mifananidzo hachisi chikumbiro chiri nyore. Manzwisisiro edu ezviono chimiro chechipiri; chinhu chatakarongerwa kuita kubva paudiki. Kubvunza zvakafanana zvemuchina haisi nzira yakatwasuka. Nechikonzero ichocho, imwe yeanonyanya kufarirwa mafomu ekuzivikanwa kweAI ndeye convolutional neural network (CNN). CNN inzira inotarisa pamapixels ari padivi peumwe neumwe. Mifananidzo iri pedyo-pedyo ndiyo inonyanya kuve ine hukama, zvinoreva kuti chinhu kana chiso chinofananidzwa nemufananidzo une kujeka.
Nepo mabhureki achitsvaga kuita mari pasocial media asi kucherechedzwa kwemufananidzo weAI kunotakura mabhenefiti akajeka, makesi ekushandisa anomhanya zvakadzika. Mota dzinozvityaira dzave kuda kuva chinhu chikuru chinotevera munyika yemotokari, uye tekinoroji yeAI yekuziva iri kubatsira kudzisimbisa. Motokari inozvityaira inokwanisa kuona zvinhu nevanhu vari munzira kuti isarovera pavari haingoitike zvoga. Inofanirwa kuziva mifananidzo kuita sarudzo ine ruzivo. Mota yega yega inozvityaira yakarongedzerwa masensor akati wandei saka inokwanisa kuona dzimwe mota dzinofamba, vatyairi vemabhasikoro, vanhu - kunyanya chero chinhu chinogona kuunza njodzi. Mota yemotokari inoda kugadzirisa njodzi dzemumugwagwa sezvinoita mutyairi ane ruzivo. Pachine zvinhu zvishoma zvekugadzirisa motokari dzinozvityaira dzisati dzapinda mumugwagwa muna 2020. Asi kana otomatiki yemotokari yapinda, kuzivikanwa kwemufananidzo weAI kuchava mumwe wevatyairi vakuru vari shure kwavo vachishanda zvakachengeteka.
⦁ Mufananidzo-kuwana
Mufananidzo wedhijitari unogadzirwa neimwe kana akati wandei ma sensors emifananidzo, ayo, kunze kwemhando dzakasiyana-siyana dzemakamera anonzwa chiedza, anosanganisira range sensors, tomography devices, radar, ultra-sonic cameras, etc. Zvichienderana nerudzi rwe sensor, inoguma data mufananidzo wakajairika we2D, vhoriyamu ye3D, kana kutevedzana kwemifananidzo. Iwo mapixel values ​​anowanzoenderana nekusimba kwechiedza mune imwe kana akati wandei spectral bhendi (mifananidzo yegrey kana mifananidzo yeruvara), asi inogona zvakare kuve ine hukama nematanho akasiyana emuviri, sekudzika, kunyura kana kuratidzwa kwemafungu esonic kana emagetsi, kana nyukireya yemagineti resonance.
⦁ Pre-processing:
Pamberi penzira yekuona komputa isati yashandiswa kudhata yemufananidzo kuitira kuburitsa imwe chidimbu cheruzivo, zvinowanzodikanwa kugadzirisa data racho kuitira kuti ive nechokwadi chekuti rinogutsa mamwe fungidziro anotaurwa nenzira. Mienzaniso ndiyo
1. Re-sampling kuitira kuvimbisa kuti mufananidzo wekubatanidza system ndeyechokwadi.
2. Kuderedza ruzha kuitira kuvimbisa kuti ruzha rwe sensor haruunzi ruzivo rwenhema.
3. Kuwedzeredza kusiyanisa kuvimbisa kuti ruzivo rwakakodzera runogona kuwanikwa.
4. Chikero-nzvimbo inomiririra kuwedzera zvimiro zvemufananidzo pazvikero zvakakodzera munharaunda.
⦁ Kubvisa zvinhu:
Mamiriro emifananidzo pamatanho akasiyana ekuomarara anotorwa kubva pamifananidzo data. Mienzaniso yakajairika yezvimiro zvakadaro mitsara, mipendero uye mitsetse
Nzvimbo dzekufarira dzenzvimbo dzakadai semakona, mabhurobhu kana mapoinzi. Zvimwe zvinhu zvakaoma kunzwisisa zvinogona kunge zvine chekuita nekugadzirwa, chimiro kana kufamba.
⦁ Kuonekwa/kukamurwa:
Pane imwe nguva mukugadzirisa sarudzo inoitwa pamusoro pekuti ndeapi mapoinzi emifananidzo kana matunhu emufananidzo akakodzera kuenderera mberi nekugadzirisa. Mienzaniso ndiyo
1. Kusarudzwa kweseti chaiyo yezvibodzwa
2. Kupatsanurwa kwenzvimbo imwe kana yakawanda yemufananidzo ine chimwe chinhu chaunofarira.
⦁ High-level processing:
Panhanho iyi inopinza inowanzova diki seti yedata, semuenzaniso seti yemapoinzi kana animage dunhu rinofungidzirwa kuti rine chimwe chinhu. Iyo yakasara yekugadzirisa inobata ne, semuenzaniso:
1. Kuona kuti data inogutsa modhi-yakavakirwa uye mashandisirwo ekufungidzira.
2. Estimation yemashandisirwo chaiwo paramita, senge chinhu chakamira kana chiyero chechinhu.
3. Kuronga chinhu chakaonekwa muzvikamu zvakasiyana.Saka, kugadzirisa mifananidzo kunobatsira AI kuziva mufananidzo uye kupindura maererano nekuzivikanwa kwemufananidzo.

Ramangwana risina musono remifananidzo

Sezvo tekinoroji inovandudza, kucherechedzwa kwemufananidzo kunodzosera zvakatokura mhedzisiro. Musoro weMachina Kudzidza kuLobster, Vladimir Pavlov anoti, "Nheyo yemasvomhu yekuzivikanwa kwechinhu yave iripo kwenguva yakareba, asi mikana yetekinoroji yekushandisa komputa kuona algorithms yakaonekwa munguva pfupi yapfuura. Nechekare, neural network inobvumira kugadzira ma detectors akakwana anokwanisa kushanda zvirinani kupfuura vanhu. A hombe jerk inobata kuvapo kweakamisikidzwa mifananidzo dataset yekudzidziswa, asi munguva pfupi iri kutevera, izvi hazvizove dambudziko. Makombiyuta ekuona mainjiniya ari kushanda nesimba pakuzvidzidzira algorithms”.Neramangwana rakanyanya kukanganiswa nekuona kutaurirana, kucherechedzwa kwemifananidzo ndiko kuchava chinhu chakakosha kumashure kwemifananidzo yakawanda yatinoona. Zvese muhupenyu chaihwo uye online.