ControlNet
ControlNet bụ ihe mgbakwunye na-enye ụdị-ọgbọ oyiyi nhazi nhazi nhazi nke ọma, na-ahapụ gị ka ị jiri akụkụ, poses, maapụ miri emi, ma ọ bụ akwụkwọ edemede mee ihe.
Nchịkọta
It turns text-to-image from a slot machine into a controllable design tool.
Ime miri emi
N'ịbụ nke Lvmin Zhang na ndị ọrụ ibe webatara na 2023, ControlNet na-ejikọta ya na ụdị mgbasa ozi a zụrụ azụ dị ka Stable Diffusion na-enweghị ịzụghachi ihe niile. Ọ na-emechi ngọngọ nzuzo nke U-Net mgbasa ozi ka ọ bụrụ nnomi enwere ike ịzụrụ, wee jikọọ nnomi ahụ laghachi azụ na nke mbụ oyi kpọnwụrụ site na oyiri-ntụgharị uche efu (zero-convs). Zero-convs ndị a na-amalite na-enweghị mmetụta, yabụ ọzụzụ na-amalite site na omume mbụ nke ụdị wee mụta ịgbanye ntụ oyi. Ndokwa ahụ bụ maapụ oghere: onyonyo ihu Canny, skeleton OpenPose, maapụ miri emi, nkpuchi nkebi, ma ọ bụ eserese siri ike. Ihe si na ya pụta bụ na onyonyo ewepụtara na-agbaso nhazi nke maapụ njikwa ebe ederede ozugbo na-edobe ụdị na ọdịnaya, na-enye ndị na-ese ihe nwere ntụkwasị obi, nhazi usoro ọzọ.
Nghọta nka nka
Ihe na-akọwa aghụghọ bụ zero-convolution. N'ihi na a na-ebunye akwa njikọ ahụ ka ọ bụrụ ihe efu efu, ngalaba ControlNet anaghị agbakwụnye ihe ọ bụla na mbụ, ya mere ihe nlereanya ahụ yiri nke mbụ na mmalite ọzụzụ. Nke a na-egbochi mkpọtụ na-emerụ ahụ nke ọkwa dị ọhụrụ ga-agbanye ma na-eme ka nhazi nke ọma guzosie ike ọbụna na obere datasets. Gradients na-abanye na zero-convs wee jiri nwayọọ nwayọọ mepee ụzọ ntụzịaka, na-amụta njikwa nhazi ahụ n'enweghị nsogbu.
Mmetụta atụmatụ
Ọsọ na ọnụ ọgụgụ
Visual AI nwere ike megharịa nyocha, nchọpụta na mkpado ọrụ n'ọtụtụ.
Mee nhọrọ
Otu ndị na-emepụta ihe nwere ike imepụta echiche ngwa ngwa site na ngbanwe akwụkwọ ntuziaka ole na ole.
Team na usoro ọrụ
Ọrụ nwere ike iji onyonyo na akara vidiyo siri ike ịhazi.
Ọdịnihu nke ControlNet
Ndokwa n'ụdị ControlNet na-aghọ akụrụngwa ọkọlọtọ na ngwaọrụ okike, yana stacking multi-condition (jikọta pose gbakwunyere omimi gbakwunyere ọnụ) yana ihe nkwụnye ọkụ dị ka T2I-Adapter na IP-Adapter. Na-atụ anya ntinye siri ike na mgbasa ozi vidiyo maka njikwa ngagharị na-agbanwe agbanwe, ndezi mkparịta ụka n'ezie, yana ụdị ejikọtara ọnụ na-anabata ọtụtụ ụdị njikwa n'otu oge, na-eme ka ahịrị dị n'etiti eserese na nsụgharị ikpeazụ.
Mmejuputa n'ezie n'ụwa
Iji skeleton OpenPose kpochie ọnọdụ agwa kpọmkwem mgbe ị na-agbanwe uwe na ndabere site na ngwa ngwa
Iji maapụ ihu Canny rụgharịa foto ụlọ ka ị na-echekwa ahịrị ụkpụrụ ụlọ ya
Na-atụgharị ihe odide aka siri ike ka ọ bụrụ ihe atụ a na-egbu maramara maka nka echiche na bọọdụ akụkọ
Itinye maapụ omimi ya mere ihe onyonyo ewepụtara na-akwanyere okirikiri nhọrọ ukwuu 3D maka ndị nrụpụta ngwaahịa na ihe njakịrị ime ụlọ.
Ihe ize ndụ & okporo ụzọ nche
Ikike onyonyo na nkwenye nwere ike bụrụ ihe egwu dị n'iwu ma ọ bụrụ na edoghị anya.
Ọrụ nlereanya nwere ike ịdịgasị iche n'ofe ọkụ, igwe mmadụ, na gburugburu.
Enwere ike ghara ịhụ ihe dị mma ma ọ bụrụ na enyochaghị oke ntụkwasị obi.
Map mmejuputa
Kọwaa ụkpụrụ nnabata maka nkenke, icheta, na ụgwọ njehie.
Nwalee na data dabara na ọnọdụ mmepụta n'ezie.
Tinye nyocha mmadụ maka obere obi ike ma ọ bụ amụma mmetụta dị elu.
Sochie ihe nlere anya wee megharịa ka emechara mgbanwe igwefoto ma ọ bụ dataset.
Nọgide na-eme nchọpụta
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 ControlNet 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
Ntuziaka na-esote
Anya SLAM
Ajụjụ a na-ajụkarị
What is ControlNet?
ControlNet bụ ihe mgbakwunye na-enye ụdị-ọgbọ oyiyi nhazi nhazi nhazi nke ọma, na-ahapụ gị ka ị jiri akụkụ, poses, maapụ miri emi, ma ọ bụ akwụkwọ edemede mee ihe. Ọ na-atụgharị ederede-na-oyiyi site na igwe oghere ka ọ bụrụ ngwá ọrụ imewe nwere ike ịchịkwa.
Kedu nsogbu ControlNet na-edozi maka ịmepụta ihe oyiyi?
ControlNet na-ahapụ ndị ọrụ na-eduzi mmepụta na ọnọdụ oghere dị ka akụkụ, poses, ma ọ bụ omimi, na-eme ka ọgbọ na-achịkwa kama ọ bụghị usoro.
Kedu ọrụ nke 'zero-convolution' layers na ControlNet?
Mgbagwoju anya mmalite efu na-enyeghị ihe ọ bụla na mbụ, yabụ ihe nlereanya ahụ dabara na nke mbụ wee mụta nchikota nke nta nke nta na kwụsie ike.
Kedu n'ime ihe ndị a bụ ntinye ntụnye ControlNet dị mma?
ControlNet na-eji maapụ oghere dị ka ọkpụkpụ pose, maapụ omimi, akụkụ Canny, na ihe mkpuchi nkewa dị ka ntuziaka nhazi.
Kedu ka ControlNet siri metụta ụdị mgbasa ozi dị ka Stable Diffusion?
ControlNet na-emechi ma na-azụ otu nnomi koodu ka ọ na-edobe U-Net izizi, na-echekwa ihe ọmụma ọ mụtara.
N'ime usoro ọrụ ControlNet, kedu ihe na-edobe ụdị na ọdịnaya mgbe maapụ njikwa na-ahazi nhazi?
Ngwa ngwa ederede na-achịkwa ụdị na isiokwu, ebe maapụ njikwa na-akwado nhazi oghere.