Ukuhlaziywa Kwesithombe: Iyini Futhi Isetshenziswa Kanjani

ngo kim Kubuyekezwe: June 29, 2022

Ukuhlaziya isithombe kuyithuluzi elinamandla lokukhipha ulwazi ezithombeni.

Lokhu kungasetshenziswa ngezinjongo ezihlukahlukene, kusukela ekulandeleni izinguquko endaweni ezungezile kuya ekubonweni kobuso kuya ekuhlaziyweni kwezithombe zezokwelapha.

Kulesi sihloko, sizokunikeza isingeniso esifushane sokuhlaziya isithombe nokuthi ungasisebenzisa kanjani. Sizofaka incazelo yokuhlaziywa kwesithombe, amasu ahlukene namathuluzi atholakalayo, kanye nokusetshenziswa okunamandla kokuhlaziywa kwesithombe.

Incazelo yokuhlaziywa kwesithombe

Ukuhlaziywa kwesithombe kuyithuluzi elisetshenziselwa ukuthola ukuqonda digital izithombe. Isetshenziswa kakhulu emikhakheni ehlukahlukene efana nezokwelapha, ucwaningo lwesayensi, ukubhekwa, ukucutshungulwa kwezithombe, nokunye okuningi.

Empeleni, kuyinqubo yokuqoqa idatha kusuka esithombeni nokuhlaziya idatha ukuze uthole ulwazi mayelana nokuqukethwe kwesithombe. Ukuhlaziywa kwesithombe kusiza ukunikeza ulwazi mayelana nekhwalithi noma izici zesithombe njengosayizi waso, umumo, ukuqina/ukukhanya, umbala ukwakheka noma ipulani yaphansi.

Kunamasu amaningana asetshenziswa ekuhlaziyweni kwesithombe njengamaqhinga okuguqula indawo njengokukhipha isithombe esingaphansi kanye nokucubungula ibhulokhi; amasu okukhipha isici athola amaphuzu noma izifunda ezithakaselwayo kusetshenziswa izitholi ezisemaphethelweni noma ama-algorithms okulandelela umngcele; amasu okuhlukanisa ahlukanisa izinto zibe izigaba; amasu okuhlukanisa ahlukanisa izinto ukusuka ngemuva; kanye nama-algorithms okuqaphela iphethini abona izinto kusetshenziswa izindlela zezibalo.

Ngokusebenzisa amathuluzi okuhlaziya izithombe ezinhlelweni zokusebenza ezifana nokutholwa kwento nokubonwa kobuso, kungathuthukisa ulwazi lomsebenzisi uma kusetshenziswa imidiya yedijithali. Kungenzeka futhi ukusebenzisa ukuhlaziya ocwaningweni lwesayensi ukuze kuhlolwe ukuhlobana phakathi kwemingcele echaziwe phakathi kwesithombe esisodwa noma iqembu. Okokugcina, lobu buchwepheshe buphinde busetshenziswe kakhulu ochwepheshe bezokwelapha kuzo zonke izici ezikhethekile ezihlanganisa i-radiology ne-pathology ukuze kuhlonzwe izimo ezihlukahlukene ngokusekelwe ezithombeni eziskeniwe ezithathwe ezigulini.

Izinhlobo zokuhlaziya izithombe

Ukuhlaziywa kwesithombe wukukhishwa kolwazi oluphusile oluvela ezithombeni zedijithali. Igama elibanzi kakhulu elihlanganisa izinhlobo eziningi zemisebenzi ehlobene nesithombe, okuhlanganisa ukuqashelwa kwento nokuhlonza, ukukalwa komumo, ukubonwa kwephethini, ukuhlaziya indawo kanye nokuhlukanisa.

Kunezinhlobo eziningi zamasu okuhlaziya izithombe; ezinye ezivame kakhulu zihlanganisa:
-Pixel-okusekelwe/Ukuhlaziywa Kwendawo - Lolu hlobo lokuhlaziya lubandakanya ukubala mathupha amaphikseli ngamanye noma amaqoqo amaphikseli ukukala izindawo (usayizi, umumo) kanye nokuminyana (ukusabalalisa).
-Ukuhlaziywa Kwezibalo - Le ndlela isebenzisa izindlela zezibalo ukuze ihlole izici ezihlobene nezifunda ezingaphakathi kwesithombe njengokugqama noma ukukhanya.
-Ukufunda Ngomshini/Ubuhlakani Obuzenzile - Ama-algorithms obuhlakani bokwenziwa (AI) anika amandla imisebenzi enamandla yokucubungula izithombe ezenzakalelayo njengokutholwa kwento noma ukuhlukaniswa kwe-semantic kusetshenziswa amanethiwekhi e-convolutional neural (CNNs).
-Ukuhlaziya Ukwakheka - Le nqubo ikala ukuthi ama-elementi angaphakathi kwesithombe asatshalaliswa kanjani ngokuhlobene ngokokwakheka—ukushintshashintsha kwamaleveli ampunga noma imibala endaweni enkulu.
-Ukuhlaziywa kwe-Histogram - Ukuhlaziywa kwe-Histogram ngokuvamile kuhilela ukulinganisa ukusatshalaliswa kwezibalo ngaphakathi kwesithombe esifana nokugqama, ukukhanya, umbala nokugcwaliswa kwesikhala.
-Izindlela Zokuhlunga - Izindlela zokuhlunga zibandakanya ukusebenza kwezibalo kuwo wonke amalungu afanayo okuqina okucutshunguliwe aguqula izingxenye ezithile zamalungu afanayo ngenkathi kugcinwa ezinye. Izihlungi zivame ukusetshenziselwa ukuthuthukisa imiphetho noma ukususa umsindo esithombeni ngokushelela kulezi zindawo ezingafuneki.
-Izindlela Zokuguqula Ezine - Le ndlela isebenzisa ukuguqulwa kwe-Fourier okubolisa izithombe zibe izingxenye ezimbalwa ngokusekelwe kububanzi befrikhwensi kanye/noma umumo okhona kuhlaka lokufaka. Lokhu kuvumela ukuhlukanisa izingxenye nezinto ezingase zibe nezici ezithile ezihlobene nezinye izingxenye ukubukeka kukonke kuhlaka - njengokwakheka noma umehluko womumo - okuvumela ukutholwa kwento kube lula phakathi nezinqubo zokuhlukanisa .

Izinzuzo Zokuhlaziywa Kwezithombe

Ukuhlaziywa kwesithombe kuyithuluzi eliyigugu lokuthola imininingwane ngezithombe zedijithali. Isetshenziselwa ukubona amaphethini, ukuhlonza izinto, nokukhipha ulwazi olubalulekile kwabezindaba. Kusukela ocwaningweni lwezokwelapha kuya ekubonweni kobuso, ukuhlaziya izithombe kungazuzisa izinhlobonhlobo zezimboni. Lapha sizohlola izinzuzo zokuhlaziywa kwesithombe nokuthi singasetshenziswa kanjani ukuthuthukisa imikhakha ehlukahlukene.

Isipiliyoni sekhasimende esenziwe ngcono

Ukuhlaziywa kwesithombe kungasiza ekuthuthukiseni ukuzizwisa kwamakhasimende, okuvumela izinkampani ukuthi ziqonde okuthandwa ngamakhasimende ngokushesha nangokunembile. Ngokusebenzisa amasu okuhlaziya izithombe, amabhizinisi angasebenzisa idatha evela ezithombeni ezifana nokubalwa kwabantu, izindawo, nezimpendulo zemizwa ukuze aqondise kangcono ukukhangisa kwawo noma okuqukethwe kube ngokwakho kumakhasimende. Ukusebenzisa le datha kungaholela ekuhlanganyeleni okuthuthukisiwe namakhasimende kanye nokwanda kokuthengisa kwenkampani.

Ukuhlaziya izithombe kusiza amabhizinisi aqonde kangcono umongo wezithombe zawo. Isibonelo, ubuchwepheshe bokubona imibala buvumela izinkampani ukuthi zibone imibala esithombeni futhi zinqume ukuthi ihlotshaniswa kanjani nemikhiqizo noma amasevisi. Lolu lwazi lungasiza amabhizinisi azuze ukuqonda ekuziphatheni kwamakhasimende mayelana nesevisi noma umkhiqizo othile. Ukwengeza, amabhizinisi angasebenzisa ubuchwepheshe bokubona izinto ukuze abone izinto ezisesithombeni futhi aqonde kangcono okuthandwa ngamakhasimende ngokusekelwe ezintweni abazithandayo.

Okokugcina, ukuhlaziywa kwesithombe kuvumela amabhizinisi ukuthi aqoqe idatha ngokushesha okukhulu kuyilapho egcina ukunemba enqubweni. Izinkampani azisadingi ukubuyekeza mathupha amakhulu ezithombe uma zizama ukunquma okuthandwa ngamakhasimende; kunalokho bayakwazi ukwenza le nqubo ngokuzenzakalelayo besebenzisa amasu okuhlaziya izithombe athuthukisa ukusebenza kahle kakhulu kuyilapho besathola imiphumela enembile ekuhlaziyeni kwayo izithombe. Lokhu kusho ukuthi amakhasimende athola ukufinyelela ekukhangiseni komuntu siqu ngokushesha kunangaphambili, okwenza izinkampani zinikeze ulwazi olungcono lomsebenzisi kumakhasimende azo.

Ukwenyuka kokunemba nokusebenza kahle

Ukuhlaziywa kwezithombe kunikeza izinzuzo eziningi zamabhizinisi, okuhlanganisa ukunemba okukhulayo nokusebenza kahle. Ngokuhlaziya izithombe esikhundleni sokuthembela ezindleleni ezisetshenziswa ngesandla noma ezisekelwe emibhalweni, ukuhlaziywa kwesithombe kunganciphisa kakhulu amaphutha okufaka idatha futhi kusheshise isikhathi sokucubungula umsebenzi othile.

Ukuhlaziywa kwesithombe kuphinde kuqede isidingo sokuhluza inani elikhulu lolwazi ngesandla. Kunalokho, i-algorithm ingathola ngokushesha idatha efanele futhi ikhombe amaphethini ukuze ahunyushwe kalula abantu. Lokhu kusheshisa inqubo yokuthatha izinqumo futhi kunciphisa amaphutha abizayo angaba khona ngenxa yephutha lomuntu noma ukugada.

Ngaphezu kwalokho, ukuhlaziya isithombe kwenza kube lula ukuhlonza amaphethini ngokushesha kumadathasethi amakhulu ngendlela emangalisayo okungathatha izinkulungwane zamahora omuntu ukuwahlunga mathupha. Lokhu kunciphisa kakhulu izindleko zabasebenzi ngenkathi kukhulisa ukunemba nokusebenza kahle kanyekanye. Ukuhlaziywa kwesithombe kuphinde kuvumele izinhlangano ukuthi zilinganise amathrendi okungenzeka ukuthi ziwagejile ngezindlela ezivamile, ukuzisiza zenze izinqumo ezingcono ngokushesha kunangaphambili.
Ngokusebenzisa ubuchwepheshe bokuhlaziya izithombe, amabhizinisi anganciphisa amaphutha okufaka idatha futhi anciphise izindleko zabasebenzi ngenkathi ekhulisa ukunemba nokusebenza kahle ekusebenzeni kwawo. Lokhu kuholela emandleni okuthatha izinqumo athuthukisiwe kanye nemiphumela enembe kakhudlwana etholakala kumadathasethi amakhulu kakhulu kunalokho okungazuzwa kusetshenziswa izindlela ezivamile zodwa.

Ukwenza izinqumo ezithuthukisiwe

Ukusetshenziswa kokuhlaziywa kwezithombe emabhizinisini ukusiza ekuthathweni kwezinqumo kungaholela ekusebenzeni okuthuthukisiwe futhi kusize izinhlangano zilawule kangcono izinsiza zazo. Ngamasu nobuchwepheshe obuhlukahlukene obuhilelekile, inhlangano inamandla okucubungula nokuhlaziya izithombe ezivela emithonjeni ehlukene ngokushesha. Lokhu kubasiza ukuthi benze izinqumo eziphusile cishe ngaso leso sikhathi.

Ukuhlaziywa kwesithombe kungasiza izinhlangano zihlele, zikhombe, futhi zihlole idatha ukuze zibeke phambili imisebenzi noma imigomo yamasu. Iphinde ivumele ukuhlaziya okusheshayo kwamaphethini ayinkimbinkimbi ngaphakathi kwezithombe, okonga isikhathi esibalulekile emisebenzini yokwenza izinqumo yezinga eliphezulu. Ukwengeza, ukuhlaziya izithombe kungasiza izinkampani zithole amathrendi noma izinkinga eziphindelelayo okungenzeka ukuthi azizange zibonwe.

Izinhlangano ziya ngokuya zisebenzisa amandla obuchwepheshe bokuhlaziya izithombe obufana nobuhlakani bokwenziwa (AI) nezinhlelo zokusebenza zokufunda komshini (ML) ukuze zenze amamodeli aqagelayo azinikeza ukuqonda kokuthi kuzokwenzekani mayelana nokuziphatha kwamakhasimende nezimo zemakethe esikhathini esizayo. Kuleli qophelo, amaqembu ezibalo zemakethe ahlala askena ngezithombe aziqoqile emithonjeni efana namawebhusayithi aqhudelana naye kanye namashaneli edijithali. Ngamathuluzi alungile, lawa maqembu angakwazi ukunquma ngokushesha ukuthi amakhasimende asebenzisana kanjani nemikhiqizo noma amasevisi enhlangano ukuze akhe imininingwane ebambekayo enikeza umkhawulo wokuncintisana ngaphezu kwabanye embonini.

Ukuhlaziywa kwezithombe kudlala indima ebalulekile ekusebenzeni kwebhizinisi namuhla ngokunikeza izinhlangano ukufinyelela emininingwaneni ebalulekile eqhutshwa yidatha ekhuthaza ukukhula nokuthuthuka ngokushesha kakhulu kunangaphambili.

Indlela Yokusebenzisa Ukuhlaziywa Kwezithombe

Ukuhlaziywa kwesithombe kuyindlela ephumelelayo yokucubungula inani elikhulu ledatha yesithombe futhi kukhishwe ulwazi oluwusizo kuyo. Ukuhlaziywa kwesithombe kuhilela ukusetshenziswa kwama-algorithms ekhompuyutha ukuze kutholwe incazelo yesithombe. Ingasetshenziselwa imisebenzi eyahlukene efana nokubonwa kwento, ukubona ubuso, nokuhlukaniswa kwesithombe. Kulesi sigaba, sizoxoxa ngokuthi ukuhlaziywa kwesithombe kungasetshenziswa kanjani kanye nezinye zezinhlelo zokusebenza.

Ukuhlukaniswa kwesithombe

Ukuhlukaniswa kwesithombe kuyigatsha lokucubungula isithombe lapho isithombe sihlukaniswa saba amasegimenti amaningi, noma izifunda, ezihlanganisa sonke isithombe. Isegimenti ngayinye imele indawo enentshisekelo kuyo ngaphakathi kwesithombe, futhi ingacatshangwa njengento esobala iyodwa. Umgomo wokuhlukanisa uwukwenza lula kanye/noma ukuguqula ukumelwa kwesithombe kube into enengqondo nelula ukuyihlaziya. Amasu okuhlukanisa asetshenziswa ekucabangeni kwezokwelapha emisebenzini ehlukahlukene, kusukela ekuboneni izakhiwo ze-anatomical kuya ekuqapheni ukuqhubeka kwesifo. Ngaphezu kwalokho, ama-algorithms okuhlukanisa ayizingxenye ezibalulekile ezinhlelweni zokugada ezizenzakalelayo futhi angasetshenziselwa imisebenzi yokuzulazula yerobhothi kanye nokutholwa kwento.

Ama-algorithms okuhlukanisa ngokuvamile ancike kuzindlela ezisuselwe kumaphikseli noma ezindleleni ezisuselwe ekuhlukaniseni. Izindlela ezisuselwe ku-Pixel zisebenzisa izinkomba ezifana nombala noma ukuthungwa ukuze zihlukanise isithombe sibe amasegimenti ahlukene. Kungenjalo, izindlela ezisuselwe ekuhlukaniseni zihlanganisa izigaba ezisekelwe kuzici ezichazayo njengomumo noma ukuthungwa ukuze kuqoqwe ngokuzenzakalelayo amaphikseli abe amasegimenti/izinto ezifanele.

Ukuhlukaniswa okusekelwe ku-Pixel ngokuvamile kuhloswe ukuhlukanisa izinto ngobukhulu bazo noma amanani wokuqhathanisa ngokubala izibalo zomhlaba wonke (isb., inani elimaphakathi) endaweni yonke. Izindlela ezisuselwe ekuhlukaniseni ngokuvamile zidinga ukufakwa ngesandla nokulebula ngaphambi kokuba i-algorithm iqale ukusebenza; Nokho lezi zindlela zinamandla kakhulu uma usebenzisana namasethi edatha amakhulu noma izinhlobo zedatha ezingaziwa ezingakwazi ukuhlukaniswa ngempumelelo kusetshenziswa amanani ephikseli kuphela.

Ukuhlukaniswa kwesithombe

Ukuhlukaniswa kwesithombe kuyinqubo yokusebenzisa ama-algorithms ekhompuyutha ukubikezela nokulebula amakilasi esithombeni esinikeziwe. Kule nqubo, kuvamile ukuba "neqoqo" lezithombe zokuqeqeshwa ezinelebula ezisebenza njengokufaka ku-algorithm yokufunda. Ngokusekelwe kulezi zithombe zokuqeqesha, izici ezifana nosayizi, umumo nombala zikhishwa esithombeni futhi zisetshenziselwa ukuhlukaniswa kwezithombe. Le nqubo ivumela imiphumela enembe kakhulu uma iqhathaniswa nokulebula okwenziwa ngesandla, lapho abantu bengenza khona amaphutha ngenxa yokungahlelwa kahle ngokwezigaba noma amaphutha okulebula.

Njengoba kunikezwe isithombe esingaziwa, umuntu angasebenzisa isihlukanisi esikhona esivele sinikeze lokhu okokufaka okuthile isigaba sokuphumayo ngokuya ngenani elichazwe ngaphambilini lamakilasi ahlukile. Okunye ukusetshenziswa okuyisibonelo kwalokhu kusuka kumasistimu okushayela imoto azimele okukunikeza amandla okubona into kuye ekutholweni kobuso ezinhlelweni zokusebenza ze-biometrics. Ukwengeza, ukuba nedatha enelebula enembile ngokuhlukaniswa kwezithombe kungasinikeza ukuqonda okwengeziwe ngendawo yethu ngokusebenzisa amadathasethi amakhulu kahulumeni ukuze ahlaziywe okwengeziwe - ikakhulukazi ekusebenzelaneni phakathi kwezinhlobo zezinto nezinhlelo zokusebenza zebhizinisi ezifana nokuhlaziya okuthengiswayo noma ukuhlaziya ukuzijabulisa.

Kusukela ekusebenziseni amanethiwekhi e-neural ngamasu okuthuthukisa i-GPU ukuze kutholwe ukucutshungulwa okufanayo kuze kufike ezinhlotsheni ezihlukahlukene zokufunda ngomshini - kunezindlela ezingenakubalwa umuntu angaqeqesha ngazo amamodeli e-AI noma asebenzise amasu okufunda ajulile emkhakheni Wokuhlukaniswa Kwezithombe; nakuba ukufunda okungagadiwe kusenezithiyo ezithile ezigcina abacwaningi ekuzisebenziseni ngokugcwele ngamandla amakhulu, lezi zindlela zisahlolisiswa futhi zicwaningwe ngenkuthalo. Ngakho-ke ukuqonda ukuthi zisetshenziswa kanjani kudinga ukuqonda okuphelele kwe-algorithms yombono wekhompiyutha kanye nolwazi oluqinile lwezobuchwepheshe olubandakanya izinqubo zobunjiniyela bamuva njengezilimi zokubhala njengePython ngezinjongo zokuqalisa kanye nokuba nesizinda esiseduze mayelana nobuchwepheshe obusafufusa obufana ne-blockchain noma ingqalasizinda engaphansi kweseva. ukuqaliswa

Ukutholwa kwento

Ukutholwa kwento ubuchwepheshe obusekelwe kukhompuyutha kanye nendlela yokuqinisekisa ukunemba ekuhlaziyweni kwedathasethi yesithombe. Lobu buchwepheshe busebenzisa ama-algorithms ahlukahlukene ayinkimbinkimbi ukukhomba izinto nezici zazo ezithombeni ezibonakala kalula. Idlala indima enkulu ezindaweni eziningi, njenge-forensics, isayensi yezokwelapha, i-automation yezimboni, ukuhlonza ubuso kanye nezinhlelo zokuhlola ezizenzakalelayo.

Ukutholwa kwento kuhlanganisa ukuhlaziya noma ukuqoqa idatha ukuze kutholwe usayizi, umumo nezinye izici zezinto ezithile ezisesithombeni. Lolu hlobo lokuhlaziywa kwesithombe lungabandakanya ukuhlonza izinto ngokusekelwe kumbala wazo, ukuthungwa noma ngisho nokuma ezikudalayo uma zihlanganiswe ndawonye. Ukusebenzisa amamodeli ezibalo nezindlela zezibalo kwenza umsebenzi ube lula kakhulu kumakhompyutha namuhla.

Inhloso eyinhloko lapha ukuhlonza ngokunembile into ngayinye evela kudathasethi yesithombe ngokusebenzisa amalebula esibaluli ayabelwe. I-algorithm idinga ukuqeqeshwa kusengaphambili ngedatha eningi, okuhlanganisa izibonelo 'ezinhle' zokuqagela okunembile mayelana nezici zento ngayinye eqondiwe ngaphambi kokuthi ukuqeqeshwa kuqale. Ngemuva kokuthi ukuqeqeshwa sekuqediwe futhi ukubikezela kube yimpumelelo, kuzolandelwa izigaba ezifana nokucatshangelwa lapho umphumela olindelekile wokufakwayo okunikeziwe usebenza nemiphumela etholwe ezifundweni zangaphambili.

Ubuchwepheshe obunjengokufunda okujulile (i-DL) ama-algorithms asiza ngokuvumela izendlalelo eziningi ezisebenza ndawonye ukuze kutholwe imiphumela enembe kakhulu phakathi nesikhathi esifushane kunendlela evamile engayinikeza ngaphambi kokuba itholakale ngaphakathi kwalolu hlobo lokucutshungulwa kwezingxenyekazi zekhompuyutha ezisetshenziswa namuhla. Ekugcineni lokhu kwenza izindlela ezisizwa yi-AI ezifana ne-Deep Learning zikhange kakhulu kunezindlela ezijwayelekile njengoba zinciphisa kakhulu amaphutha ngenkathi zinikeza amamethrikhi okunemba kokutholwa kwento angakhiqiza kalula okukhiphayo nasezimeni zesikhathi sangempela.

Amathuluzi Okuhlaziya Isithombe

Ukuhlaziywa kwesithombe kuyinqubo ehilela ukukhipha ulwazi olubalulekile ezithombeni kusetshenziswa izindlela zayo zokubala. Le nqubo isetshenziswa emikhakheni eyahlukene, njengokuthwebula izithombe zezokwelapha, ubuhlakani bokwenziwa, kanye nokuzwa okukude. Ngenhlanhla, kunezinhlobonhlobo zamathuluzi atholakalayo okuhlaziya isithombe, okwenza inqubo ibe lula futhi isebenze kahle. Ake sibheke amanye amathuluzi atholakalayo okuhlaziya isithombe.

I-OpenCV

I-OpenCV iwumtapo wezincwadi wemisebenzi yokuhlela ehloselwe kakhulu ukubona ikhompyutha yesikhathi sangempela. Isekela amawindi ahlukahlukene, i-Linux, iplatifomu nezinye izinkundla zesoftware ngakho-ke ivame ukusetshenziselwa ukuhlaziya izithombe nemisebenzi yokucubungula izithombe. Izindawo zesicelo se-OpenCV zifaka: ukubona ngekhompyutha, ukubona ubuso, ukuhlonza into, ukulandelela ukunyakaza, ukuhlukaniswa nokuqashelwa.

I-OpenCV inikeza anhlobonhlobo amathuluzi anamandla angasetshenziswa ukwenza imisebenzi eyinkimbinkimbi njengokutholwa kwento, ukubonwa kobuso, ukukhipha isici okuzenzakalelayo nokunye. Inomphakathi osebenzayo ohlinzeka ngezixazululo ezisebenziseka kalula ezinkingeni ezijwayelekile. Umtapo wezincwadi uklanyelwe ukuhlanganisa nezilimi eziningi zokuhlela ezifana ne-C++, i-Java noma i-Python. Umtapo wolwazi wenza abathuthukisi bakwazi ukusebenza ngezithombe ngezilimi abazifisayo ngaleyo ndlela banciphise isikhathi sokuthuthukisa kakhulu. I-OpenCV isekela noma yiluphi uhlobo lwedatha (njengezithombe eziku-2D nokwakheka kwe-3D) okuyenza ibe yithuluzi elisebenza ngezindlela eziningi phakathi kwayo yonke imitapo yolwazi yokucubungula izithombe etholakalayo namuhla.

Amathuluzi afakwe ku-OpenCV yilawa:
-Image Ukuhlunga
-Ukuhlukaniswa kwento
-Izici isizinda
-Ukumatanisa nokuhlukaniswa kwesici
-Ukulandelela into kanye nokuhlaziywa kokunyakaza
-Ukwakhiwa kabusha kwe-3D kwezinto ezifundwayo

Amathuluzi ayo asetshenziswa kabanzi ekuhloleni okuzenzakalelayo okubonakalayo kuzilungiselelo zezimboni ngenxa yezici zayo ezinenzuzo njengokusetshenziswa kwememori okuphansi, ukubulawa ngokushesha kumapulatifomu amaningi (ikakhulukazi ashunyekiwe), ukuphatheka phakathi kwezinhlelo ezisebenzisa imitapo yolwazi ejwayelekile njenge-OpenMPI/MPI4Py. Ilayisense yayo yomthombo ovulekile iyenza ithandeke nakakhulu kubasebenzisi abafuna ukuyisebenzisa ngaphandle kwezinkokhelo zenzuzo noma amalayisense avela ezinkampanini noma izinhlangano ezifana nezibalo njll...

I-TensorFlow

I-TensorFlow iyilabhulali yesofthiwe yomthombo ovulekile esetshenziselwa ukuthuthukisa nokuqeqesha amamodeli okufunda ajulile. Amasistimu okufunda ajulile, njengamamodeli okuhlaziya izithombe, akhiwe izendlalelo eziningi zama-neurons acubungula idatha yokufaka, njengesithombe. I-TensorFlow yenza abathuthukisi bakwazi ukuthuthukisa kalula futhi basebenzise amamodeli okufunda ajulile angasetshenziswa embonweni, ekucubungulweni kolimi lwemvelo nakwezinye izindawo.

Inzuzo enkulu yokusebenzisa i-TensorFlow ukuthi inikeza indlela elula yokudala nokusebenzisa amamodeli okufunda emshini anamandla ngokushesha nangempumelelo. I-TensorFlow iphinde isekele izinhlelo zokusebenza eziningi kuzo zonke i-MacOS, iWindows, i-Linux nezinkundla zeselula. Futhi ngenxa yokuthi iwumthombo ovulekile futhi inosekelo olucebile lomphakathi, ungasebenzisa amathuluzi akhona enqubweni yakho yokuthuthukisa ngaphandle kokuthi usungule kabusha amasondo noma uchithe isikhathi esengeziwe ekususeni iphutha lekhodi kusukela ekuqaleni.

Ngokungeziwe ekuqeqesheni amamodeli amasha kusukela ekuqaleni, i-TensorFlow ikuvumela ukuthi uphinde usebenzise amamodeli aqeqeshwe ngaphambilini ukuze ufunde ukudlulisa noma ukulungisa kahle amasethi ezinkinga ezithile. Lokhu kwenza kube nokwenzeka ukwakha ngokushesha izixazululo ezenziwe ngokwezifiso ezihambisana nezinkinga ezithile kuyilapho kusizakala ngolwazi olukhona olwakhelwe kumanethiwekhi akhona. Ngalokhu kuvumelana nezimo esandleni, abathuthukisi bangaqala ukuhlola izakhiwo ezihlukile ngokushesha kunangaphambili - ukuzithuthukisa futhi zisebenze ngemiphumela enembile kakhulu ngokushesha.

I-Google Cloud Vision

I-Google Cloud Vision iyithuluzi elinamandla lokucubungula izithombe elidizayinelwe ukusiza amabhizinisi ahlaziye izithombe futhi athole imininingwane ebalulekile. Isekelwe kubuchwepheshe obufanayo obusetshenziswa Izithombe ze-Google, okuyenza ithembeke futhi ibe lula ukuyisebenzisa. Leli thuluzi elisekelwe emafini livumela amabhizinisi ukuthi acubungule inani elikhulu lezithombe ngokushesha, abone ubuso, ahlonze izinto, futhi akhiphe umbhalo ngaso sonke isikhathi ngenkathi egwema umsebenzi wezandla.

Esinye isici esihle kakhulu se-Google Cloud Vision yikhono layo lokuhlonza izinto ezivela enhlobonhlobo yezigaba njengezilwane noma imikhiqizo. Ngesici "sokuthola ilogo", abasebenzisi bangaxhuma nemikhiqizo efana ne-Facebook noma i-Instagram ukuze bathole ulwazi oluningiliziwe lokulandela umkhondo nokuhlaziya. Isici "sokuthola indawo eyisizinda" sisiza abasebenzisi ukuthi babone izimpawu zendawo ezibazungezile futhi baqonde isiko labo lendawo kangcono ngokuphazima kweso.

Ngaphandle kwamakhono okubona into, i-Google Cloud Vision iphinde inikeze abasebenzisi ngokuhlaziywa kwemizwa kanye nethuluzi lamagama-ncazo elizenzakalelayo lezithombe — elilungele ukuhlaziya ngokushesha! Ukwengeza, leli thuluzi linikezela ngamakhono okubona ubuso alungele izinjongo zokumaketha njengoba ebona izici ezifana neminyaka nobulili ngokunemba okuphezulu. Okokugcina, amandla okukhipha igama elingukhiye avumela abasebenzisi ukuthi baskene ngokushesha ngezithombe zokufaka ngenkathi bethola amagama angukhiye afanelekile angasetshenziswa ezinqubweni zakamuva noma izinhlelo zokusebenza.

Sengiphetha, i-Google Cloud Vision iyinketho enhle uma ufuna ithuluzi lokucubungula izithombe ukuze usize amabhizinisi ahlaziye izithombe zawo kahle kakhulu kunangaphambili!

Isiphetho

Ukuhlaziya isithombe kuyithuluzi elinamandla lokuqoqa imininingwane ezithombeni ezidijithali. Ingasetshenziswa emikhakheni eminingi ehlukene njengemithi, izimoto, kanye nokuthatha izithombe. Izinzuzo zokusebenzisa ukuhlaziya izithombe ziningi futhi zingonga isikhathi, imali, nomzamo. Kulesi sihloko, sibheke izinto eziyisisekelo zokuhlaziywa kwesithombe, ukuthi kuyini, nokuthi kungakuzuzisa kanjani. Siphinde saxoxa ngezinhlobo ezahlukahlukene zokuhlaziya kanye nemikhuba engcono kakhulu yokuhlaziya isithombe esiyimpumelelo. Sengiphetha, kusobala ukuthi ukuhlaziya izithombe kuyithuluzi elinamandla lokuqoqa imininingwane nokwenza izinqumo eziqhutshwa idatha unolwazi.

Isifinyezo

Ukukhetha okuthosiwe kwekhofi lakho kuncike ekunambithekeni namandla owakhethayo kanye nokuncamelayo kwesifunda. Ama-roasters ahlukene anikeza izinhlobonhlobo zamagama nezigaba zemibala, kodwa ngokuvamile, ama-roast amancane azoba ne-flavour ethambile kanye nokuqukethwe kwe-caffeine ephakeme kancane kuyilapho okuthosiwe okumnyama kubaba kakhulu kune-asidi encane. Izinhlobo ezidumile zazo zonke zihlanganisa i-Light City, i-Half City, i-Cinnamon, i-City, i-American, i-Breakfast, i-Full City, i-High Continental New Orleans European Espresso Viennese Italian French. Kungakhathaliseki ukuthi ungumuntu othanda ubhontshisi wekhofi okhanyayo noma omnyama—noma phakathi nendawo—qiniseka ukuthi uyabuza ngaphambi kokuthenga ukuze uthole lokho okufunayo!

Ukufunda Okwengeziwe

Ukusetshenziswa kokuhlaziya izithombe ukuze uthole ukuqonda futhi ufinyelele iziphetho ezithombeni kuya ngokuya kuduma ezimbonini eziningi, kusukela kwezokunakekelwa kwempilo kuye kwezokuphepha. Uma ungathanda ukufunda okwengeziwe ngalobu buchwepheshe, naka lezi zinsiza ezilandelayo:

Imibhalo nezifundo
-“Ukuhlaziywa Kwesithombe: Kuyini Futhi Kusetshenziswa Kanjani” ngu-Oracle
-“Isingeniso Sokuhlaziywa Kwezithombe Kubantu Abangazi Lutho Ngokuhlaziywa Kwesithombe” ngu-Kent Woodard
-“Izinyathelo Eziyisi-8 Zokuqonda Ukuhlaziywa Kwesithombe Nokucutshungulwa Ekufundeni Ngomshini” nguVictor Charpenay
Amabhuku nezincwadi
-Ukucutshungulwa Kwesithombe: Izimiso Nezicelo nguMilan Sonka et al.
-Computer and Machine Vision – Theory, Algorithms, Practicalities (4th Ed.) ka-ER Davies
-Ukucutshungulwa Kwesithombe Sedijithali nge-OpenCV - I-Bibliography (Ver 4.1) Ihlanganiswe NguDavid Dardas Webinars & Podcasts
-Computer Vision Through Machine Learning Webinar ephethwe yi-Treehouse AI
-Izisekelo Zokufunda Ngomshini I-Podcast ephethwe i-Google Cloud Platform Kukhona nezinye izincwadi eziningi, izindatshana, ama-webinars, ama-podcasts, ama-workshops nezinye izinsiza ezitholakalayo ezingakusiza uqonde kangcono indlela yokuhlaziya isithombe kanye nendlela yokuyisebenzisela izinhlelo zokusebenza ezithile.

Sawubona, nginguKim, umama kanye nomshisekeli we-stop-motion onesizinda ekudalweni kwemidiya nokuthuthukiswa kwewebhu. Nginothando olukhulu lokudweba nokugqwayiza, futhi manje sengingena ngekhanda emhlabeni onyakazayo. Ngebhulogi yami, ngabelana ngezifundo zami nani bafana.