Ama-CNN Asekelwe esifundeni
Ama-CNN Asekelwe Kusifunda (R-CNNs) awumndeni wezitholi zezinto eziphakamisa kuqala izifunda zekhandidethi esithombeni, bese zisebenzisa i-CNN ukuze zihlukanise futhi zifake ebhokisini ngokuqondile into ngayinye.
Uhlolojikelele
They turned image classification into full object detection, locating and labeling many objects at once.
I-Deep Dive
Izimpendulo zokuhlukaniswa kwezithombe 'yini ekulesi sithombe?' kodwa ukutholwa kufanele kuphendule nokuthi 'kuphi, futhi bangaki?' I-R-CNN yasekuqaleni (2014) yasebenzisa i-algorithm yangaphandle (Usesho Olukhethiwe) ukuphakamisa izifunda ezingaba ngu-2,000, yasonteka ngasinye saba usayizi ogxilile, futhi yasebenzisa i-CNN kuzo zonke, eyayinembile kodwa ihamba kancane kabuhlungu. I-R-CNN esheshayo isheshise lokhu ngokusebenzisa i-CNN kanye kuso sonke isithombe nezici zokuhlanganisa isifunda ngasinye (i-RoI pooling). I-R-CNN eshesha kakhulu yabe isishintsha i-Selective Search yafaka i-Regional Proposal Network (RPN) efundiwe, okwenza wonke umzila wepayipi uphele futhi usondele nesikhathi sangempela. Imaski i-R-CNN iyinwebe nakakhulu ukuze ikhiphe imaski yezinga le-pixel entweni ngayinye etholiwe.
I-Technical Insight
Ukuphumelela okubalulekile okubalulekile ukuhlanganisa i-RoI: esikhundleni sokuqalisa kabusha i-CNN kuwo wonke amabhokisi ahlongozwayo, inethiwekhi ihlanganisa imephu yesici esisodwa esabelwe esithombeni, bese inqampuna futhi ishintsha usayizi wezici ngaphakathi kwesifunda ngasinye esithandwayo ukuze ibe igridi engashintshi. I-RPN ye-R-CNN eshesha kakhulu ishelela kuleso sici semephu ebikezela izikolo 'zento' nokulungiswa kwamabhokisi amabhokisi okusetwa ngaphambilini amasayizi ahlukahlukene kanye ne-aspect ratio, ekhiqiza iziphakamiso cishe mahhala.
I-Strategic Impact
Isivinini nesikali
I-Visual AI ingakwazi ukuhlola, ukutholwa, nokumaka imisebenzi esikalini.
Yakha ukukhetha
Amathimba aqanjiwe angakwazi ukulinganisa imiqondo ngokushesha ngezibuyekezo ezimbalwa ezenziwa mathupha.
Ithimba kanye nokusebenza komsebenzi
Imisebenzi ingasebenzisa amasiginali wesithombe nawevidiyo obekunzima ukuwenza ngaphambilini.
Ikusasa lama-CNN Asekelwe Esifundeni
Izithungatha ze-R-CNN ezinezigaba ezimbili zihlala ziqinile lapho ukunemba kubaluleke khona kakhulu, kodwa izitholi zesiteji esisodwa (YOLO, SSD) kanye nezitholi ezisekelwe ku-Transformer ezifana ne-DETR, ezeqa ngokuphelele amahange neziphakamiso eziklanywe ngesandla, ziya ngokuya zidume ngesivinini nobulula. Ukuthambekela kubheke ekupheleni ukuya ekupheleni, okungenakho i-ancho, ukutholwa okusekelwe embuzweni. Noma kunjalo, imibono ewumgogodla yohlu lwe-R-CNN, izici ezabiwe kanye nokucabanga kwezinga lesifunda, kuyaqhubeka nokuba nomthelela ekuhlukaniseni, ividiyo, kanye nezinhlelo zokuthola i-3D.
Ukuqaliswa Komhlaba Wangempela
Ukuthola nokubala imikhiqizo emashalofini okudayisa ukuze ulawule i-inventory
Isegmentation yamaseli noma izitho zomzimba kumaskeni wezokwelapha kusetshenziswa Imaski R-CNN
Ukuhlonza amaphutha nezindawo zawo emugqeni wokukhiqiza wefekthri
Ukuthola izimoto eziningi nabahamba ngezinyawo kuma-autonomous-driving camera feeds
Izingozi & Guardrails
Amalungelo ezithombe kanye nemvume kungaba ubungozi bezomthetho uma ukuvela kungacacile.
Ukusebenza kwemodeli kungahluka kukho konke ukukhanya, izibalo zabantu, kanye nezindawo.
Okuhle okungelona iqiniso kungase kungabonakali ngaphandle uma izinga lokuzethemba liqashelwa.
Ukuqalisa Umhlahlandlela
Chaza indlela yokwamukela yokunemba, ukukhumbula, nezindleko zamaphutha.
Hlola ngedatha efana nezimo zangempela zokukhiqiza.
Engeza isibuyekezo somuntu ukuze uthole ukuzethemba okuphansi noma izibikezelo zomthelela omkhulu.
Landelela ukukhukhuleka kwemodeli bese uqinisekisa kabusha ngemva kwezinguquko zekhamera noma zesethi yedatha.
Qhubeka Uhlole
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Region-Based CNNs quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Umhlahlandlela olandelayo
Amamodeli Okukhiqiza Asuselwa kumaphuzu
Imibuzo evame ukubuzwa
What is Region-Based CNNs?
Ama-CNN Asekelwe Kusifunda (R-CNNs) awumndeni wezitholi zezinto eziphakamisa kuqala izifunda zekhandidethi esithombeni, bese zisebenzisa i-CNN ukuze zihlukanise futhi zifake ebhokisini ngokuqondile into ngayinye. Baguqule ukuhlukaniswa kwesithombe kwaba ukutholwa kwento egcwele, ukuthola nokulebula izinto eziningi ngesikhathi esisodwa.
Uyini umbono oyinhloko we-CNN esekwe esifundeni?
Ama-R-CNN aphakamisa izifunda ezingase ziqukathe izinto, bese esebenzisa i-CNN ukuze ahlukanise isifunda ngasinye futhi acwenge ibhokisi laso elihlanganisayo.
Kungani i-R-CNN yasekuqaleni (2014) yayihamba kancane?
I-R-CNN yoqobo yaqhuba i-CNN ngokuzimela mayelana neziphakamiso zesifunda ezingaba ngu-2,000 ngesithombe ngasinye, ebesibiza kakhulu ngokubala.
Yini eyethule i-Faster R-CNN ukuze ingene esikhundleni se-Selective Search?
I-Faster R-CNN yethule i-Regional Proposal Network efundiwe, okwenza ukukhiqizwa kwesiphakamiso kube ingxenye yenethiwekhi eqeqeshekayo futhi kusheshe kakhulu.
Kufezani ukuhlanganisa i-RoI?
I-RoI yokuhlanganisa ikhiphe amagridi esici anosayizi ogxilile kusukela kumephu yesici esisodwa esabiwe endaweni ngayinye, ukugwema ukuqalwa kabusha nokusheshisa ukutholwa.
Yimaphi amandla engeziwe IMask R-CNN eyengezayo?
Imaski i-R-CNN inweba i-Faster R-CNN ngokubikezela imaski enembayo yeleveli yephikseli entweni ngayinye, ivumela ukuhlukaniswa kwesibonelo.