Riba:
- Yin amfani da hankali na wucin gadi a cikin bayanin rashin son zuciya da rukunin farko na abubuwan da aka gano, da kuma iya bambance bayanan al'adu-lokaci daga aikin ƙwararru tare da kasidar bincike.
- Ikon fahimtar yadda ingancin hoto (ma'auni, haske, katin launi) da bayanan metavisual kamar rubutu, bayanin martaba, mahallin mahallin ke tantance daidaiton rarrabuwa.
- Ƙarfin gane halin basirar wucin gadi don tilasta tsarin da aka saba da shi a cikin abubuwan da ba a saba gani ba da kuma tabbatar da kowane sifa tare da tushe mai zaman kansa.
Ɗaya daga cikin mafi gajiyarwa amma mai ba da labari ayyuka na tono shine nemo rarrabuwa (raba abubuwan da aka samo bisa ga nau'i, tsari da lokaci). Dubun yumbura, ɗaruruwan tsabar kudi, kayan aikin dutse (lithic), kashi, gilashi da abubuwan ƙarfe da aka gano a cikin yanayi dole ne a bincika kuma a yi rikodin su ɗaya bayan ɗaya. A cikin wannan naúrar, za mu koyi yadda nazarin hoto (cire fasali ta atomatik kamar siffa, launi, rubutu daga hoton abu) da rarrabuwa na tushen AI yana haɓaka aikin neman, amma me yasa yanke shawara na ƙarshe koyaushe ya kasance tare da gwani.
Babban ka'idar ita ce: AI da sauri ƙungiyoyi tare waɗanda suke kama; amma mutum na iya yin bayanin al'adu-lokaci kamar "wannan yanki shine samar da Gabashin Girka na karni na 6 BC" bisa ga kasida da ido na kwararru. AI shine mataimakan tantancewa da aunawa, ba ikon buga rubutu ba.
Matakan nemo rarrabuwa
1. Kamun gani. Kowane abin da aka samo ana ɗaukar hoto a daidaitaccen tsari, tare da sandar sikelin (mai mulkin nuna girman) da katin launi. Daidaitaccen haske da baya suna da mahimmanci don nazarin hoton AI; Mummunan hoto yana nufin mummunan rabewa.
2. Feature hakar. AI; Yana fitar da siffofi masu ma'auni kamar bayanin martaba, nau'i (rim, kasa, rike), jiyya na saman (polishing, gilashi, fenti), launi da rubutu. Kullu (nau'in yumbu da aka kora; launinsa da abubuwan da ke tattare da shi suna nuna wuri da fasaha na samarwa) shine kashin bayan ilimin ilimin archaeological.
3. Pre-group. AI yana rarraba dubban sassa zuwa gungu dangane da kamanni na gani: "wataƙila akwati iri ɗaya," "siffa iri ɗaya," "kullu iri ɗaya." Wannan yana sanya tulin tsari a gaban gwani.
4. Siffar gwani. Masanin ya danganta lokaci, wurin samarwa, da aiki ga kowane rukuni ta hanyar kasida, nau'in nau'in, da kwarewarsa. AI yana cewa "wannan shine zamanin Roman" alama ce, ba yanke shawara ba.
5. Rijista. Kowane nemo yana shigar da bayanai tare da lambar mahallinsa, ma'auni, hoto da sifa (wannan shine batun raka'a 5).
Alamomi: Ka sa AI ta “bayyana” abin da aka samo, ba “bayyana” shi ba. "Wannan wane zamani ne?" a maimakon haka tambayi "da gangan jera abubuwan da ake iya aunawa (siffa, gefen, saman, launi) na wannan ɓangaren"; Kuna yanke shawara na tsawon lokaci tare da kundin bayani. Bayanin abin dogara ne, sifa yana da haɗari.
Ya kamata a biya kulawa ta musamman ga tsabar kudi da rubutun.
Tsabar kudi suna da ban sha'awa a nazarin hoto saboda suna da yawa kuma sun daidaita. AI na iya taimakawa wajen gane hoton da ke fuskar tsabar kudin (ban da baya) da alamar da ke baya (a baya). Amma gano tsabar kuɗi yana buƙatar ƙwarewar ƙima: ko rubutun da aka sawa, rubutun jabu, ko wani nau'in da ba kasafai ba, rashin rarrabawa duk a bayyane yake. AI yana cewa "wannan shine tsabar kudin wancan sarki" tabbas yana tabbatar da daidaitaccen kundin kundin tsabar kudi. Hakanan ya shafi abubuwan da aka rubuta da hatimi.
Tsanaki: AI yana kula da "daidaita" abin da aka samo wanda yayi kama da sanannen nau'i zuwa irin wannan sanannen nau'in; Yana tilasta misalan da ba kasafai ba, na gida ko na yau da kullun cikin mafi yawan sanannun tsarin. Kasance cikin mafi kyawun shakku game da batun farko game da AI a cikin wani sabon abu da ba a saba gani ba.
uku mini lokuta
Case 1 - Pre-rarraba rage kaya. 8,400 ceramic sheds an tono su a cikin wani yanki na Zamanin ƙarfe. Ƙungiyar gani ta AI ta raba sassan zuwa kusan nau'ikan 40 na farko; Maimakon kawar da kowane yanki daga karce, gwani ya bi ta cikin gungu, kuma an rage lokacin rarrabawa da kusan kashi uku. Masanin ya raba nau'ikan kullu daban-daban guda biyu waɗanda AI ta haɗa ba daidai ba.
Case na 2 - An kama sifa ta ƙarya. Wani ma'aikaci ya gabatar da tsabar kudi ga AI; AI cikin amincewa ya dangana shi ga sanannen nau'in Roman. Lokacin da masanin numismatic ya kwatanta shi da kasidar, ya ga cewa alamar da ke baya da kuma almara (rubutu a kan tsabar kudin) tsari sun bambanta kuma wannan wani yanki ne na kwaikwayo. AI ta tilasta mafi yawan sanannun nau'in.
Case 3 - Matsalar rashin daidaituwar launi. Ƙungiya ɗaya ta ciyar da hotunan AI da aka ɗauka a ranaku daban-daban a cikin haske daban-daban; Launukan kullu sun kasance marasa daidaituwa kuma an karye rukuni. Lokacin da ƙungiyar ta canza zuwa katin launi da tsarin haske akai-akai, daidaiton haɗawa ya ƙaru sosai. Darasi: ingancin hoto shine ingancin rarrabawa.
Samfura huɗu masu kwafi
1) Bayanin tsaka tsaki:
Matsayinku: sami mataimakiyar bayanin. Zan ba ku hoto/data nemowa. Jeri kawai abubuwan OBSERVABLE da gaske: siffa (baki/ƙasa/jiki/hannu), ƙimanta girman, ƙarewar ƙasa, launi, ado, yanayin karyewa/sawu. KADA KA YI NASARA ZUWA GA LOKACI, AL'ADA KO WURAREN INGANTA. Idan ba ku da tabbas game da fasalin, rubuta "rashin tabbas".
2) Gabatarwa:
A ƙasa zan liƙa bayanan bayanin sassan [N]. Raba su zuwa rukuni bisa ga kamanni da kullu. Ga kowace ƙungiya: rubuta abubuwan gama gari, lambobin ɓangaren da suka faɗo cikin ƙungiyar, da ƙa'idodin da ke haɗa ƙungiyar. Saka kowane sassan da ba a sani ba akan jerin "don bita" daban. Kar a ba da ainihin nau'in suna.
3) Samar da lissafin ƙididdiga don ambato:
Matsayinku: Mashawarcin ƙwararrun masana'antun yumbu. Don ƙungiyar mai zuwa, jera ma'auni da nake buƙata in bincika lokacin VEROFING na lokaci/abun samarwa (binciken kullu, kasidar nuni, gano layi ɗaya, kwanan wata mahallin, da sauransu). Kada ku yi tunani; Faɗa mini abin da zan bincika.
4) Duba ingancin hoto:
Yi la'akari da waɗannan hotuna da aka samo: akwai ma'aunin ma'auni, hasken yana daidaitawa, ana iya ganin katin launi, bayanan baya a sarari, yanki a sarari? Wadanne hotuna ya kamata a sake ɗauka don rarrabawa? Jera matsaloli tare da lambar ɓangaren.
Rauni mai ƙarfi / Ƙarfi mai ƙarfi
Rawanin faɗakarwa:
Wace wayewa ce wannan tukwane, shekarunta nawa, kuma me aka yi amfani da shi?
AI yana tsara wayewar wayewa, daidaitaccen shekaru, da madaidaicin aiki daga hoto ɗaya; Duk basu da tushe kuma suna da tabbaci.
Ƙarfi mai ƙarfi:
Matsayinku: sami mataimakiyar bayanin. Bayyana abubuwan da ake iya gani na yanki kawai a cikin hoton: nau'i na nau'i (baki/ƙasa?), saman (mai kyalli/glazed, kone?), launi, ado, ƙididdiga kauri. Bayan haka, dangane da wannan bayanin, gaya mani abin da ake buƙata ƙarin bayanai (binciken kullu, zanen bayanin martaba, mahallin) don fayyace aiki da lokaci. Kada ku yi maƙasudin wayewa / shekaru daidai.
Bambanci: na farko yana buƙatar yanke shawara kuma yana karɓar dacewa; na biyu yana buƙatar kwatanci da taswirar hanya mai tabbatarwa.
Bayan na gani: kullu, mahallin da bayanin martaba
Wani rami a cikin binciken hoto na AI shine kawai yana kallon bayyanar saman. Duk da haka, sau da yawa ba a iya ganin kashin baya na nau'in kayan tarihi na kayan tarihi. Kullu (yankin da aka kora na yumbu, abubuwan da ke da shi da kuma nau'insa a cikin sashin da aka karya) yana bayyana inda kuma yadda aka samar da jirgin ruwa da kuma abin da aka yi amfani da shi; Ko da yake guda biyu suna kama da saman, idan kullunsu ya bambanta, suna iya kasancewa cikin al'adun samarwa daban-daban. Don haka, rarrabuwa mai tsanani yana buƙatar ɗaukar sashin da ya karye don ɗaukar hoto.
Hakazalika, bayanin martaba (zanen sikelin ɓangaren gefen-jiki-ƙasa na jirgin ruwa) yana ma'anar sigar sosai fiye da hoto ɗaya; kusurwar baki, kaurin gefen da siffar ƙasa sune alamun lokacin. AI na iya taimakawa ta atomatik cire zanen bayanin martaba da rukunin bayanan martaba iri ɗaya, amma ƙwararren yana tabbatar da zane da aunawa. A ƙarshe, mahallin (a cikin wane Layer aka sami yanki, tare da menene) ƙaƙƙarfan alamar soyayya mai zaman kanta ba tare da kamanni na gani ba: guda a cikin mahallin rufaffiyar sau da yawa daga lokaci guda. Idan kun ba AI kawai hoto kuma ku tsallake rubutu, bayanin martaba, da mahallin, kuna watsi da mafi kyawun shaidarku.
Tukwici: Lokacin ba da kayan tarihi ga AI, samar da ra'ayoyi guda uku idan zai yiwu: saman waje, saman ciki, da yanki mai karye. Hakanan haɗa lambar mahallin da kowane ma'aunin bayanan martaba. Ƙarin bayanan rashin son zuciya yana nufin mafi yawan abin dogara.
Teburin amfani da AI ta nau'in nemo
Nau'in samu
AI yana da ƙarfi
Haɗari / kasuwancin ɗan adam
Tabbaci na wajibi
tukwane
Form/kullu pre-grouping
Lokaci, wurin nunin samarwa
Kasidar Magana, Binciken kullu
tsabar kudi
Face/ icon pre-gane
Halin, kwaikwayi/ ganowa ba kasafai ba
Kundin tsabar kudin, gwani
Lithic (kayan aikin dutse)
Nau'i da ma'aunin girma
Aiki, fasaha review
Amfani-sawa bincike
gilashin / karfe
Launi, form pre-rabuwa
Haɗin kai, tarihin tarihi
Binciken dakin gwaje-gwaje
samu kadan
Haɗin hoto mai ƙima
Aiki, fassarar mahimmanci
Magana, misali mai kama da juna
Kuskuren gama gari
- Neman nassoshi maimakon kwatance. Tambayi AI don "abin da yake kama," ba "abin da yake" ba; Yi shawararku tare da kasida.
- Rarraba tare da mummunan hoto. Ba tare da ma'auni ba, haske da katin launi abin fitarwa ba shi da tabbas.
- Yarda da batun farko na AI game da samuwar da ba kasafai ba. AI yana tilasta tsarin al'ada cikin mafi yawan sanannun tsarin.
- Tsallake manna/ bayanan mahalli. Kamancen gani kawai yaudara ce; kullu da mahallin suna yanke hukunci.
- Kuskuren rukuni na farko don takamaiman nau'in rubutu. Rukuni farawa ne; Typology kwararre ne kuma aikin tunani.
a takaice
AI a cikin rarrabawa; Yana adana lokaci mai mahimmanci a cikin bayanin gani, haɓaka fasalin da ayyuka na farko. Amma ra'ayin al'ada-lokaci-lokaci, kwaikwayi/rarar ganowa da fassarar aiki na ƙwararru ne da kasidar bincike. Faɗa AI "bayani" ba "bayani" ba; daidaita ingancin hoto; tabbatar da kowane sifa tare da tushe masu zaman kansu; A cikin samun da ba kasafai ba, zama mafi yawan shakka game da amsar farko ta AI.
Aikin aikace-aikace
Shirya hotunan da aka samo guda 10 da kuke da su (ko zaɓi daga kundin kasida) a daidaitaccen tsari. Kwatanta kowane ɗayan tare da samfurin "Bayyanawa Tsakani", sa'an nan kuma haɗa su tare da samfurin "Pre-grouping". Sannan kwatanta ƙungiya zuwa ƙasidar tunani na gaske kuma lura inda ƙungiyar AI ta kasance tabo kuma inda ba daidai ba.
jerin abubuwan dubawa
- [ ] Na daidaita hotuna tare da ma'auni, haske da katin launi.
- [ ] Ina son bayanin tsaka tsaki daga AI, ba sifa ba.
- [ ] Na duba rukunin farko tare da ƙwararren ido da kasida.
- [ ] Na yi tambaya game da sifa ta farko ta AI a cikin abubuwan da ba safai ba / na yau da kullun.
- [ ] Na tabbatar da kowane bayanin al'ada-lokaci tare da tushe masu zaman kansu.