Netwọk ndị fọdụrụ
Netwọk ndị fọdụrụnụ (ResNets) bụ netwọkụ akwara dị omimi nke na-agbakwunye 'njikọ sụọ' na-ekwe ka akwa akwa mụta obere mgbanwe kama mgbanwe zuru oke.
Nchịkọta
This simple trick made it possible to train networks hundreds of layers deep, sparking a leap in image recognition accuracy.
Ime miri emi
Tupu ResNets, ikpokọta ọtụtụ n'ígwé mere netwọk na-eme ka ọ ka njọ, ọbụlagodi na data ọzụzụ, nsogbu a na-akpọ mmebi. N'afọ 2015, ndị nchọpụta Microsoft Kaiming He na ndị ọrụ ibe webatara ihe mgbochi ahụ: kama ịrịọ nchịkọta nke ọkwa ka ha mepụta mmepụta H (x) ozugbo, ha na-ahapụ ya ka ọ mụta ihe fọdụrụ F (x) = H (x) - x, wee tinye ntinye mbụ x azụ site na ụzọ mkpirisi. Ọ bụrụ na oyi akwa adịghị mkpa, ọ nwere ike ịmụta ime ihe ọ bụla (F(x) = 0). ResNet-152 meriri asọmpi ImageNet 2015 na njehie n'elu-5 nke ihe dị ka pasent 3.6, na-eti atụmatụ ọkwa mmadụ, na ihe owuwu ya ghọrọ ọkpụkpụ ndabere maka nchọpụta, ngalaba, na onyonyo ahụike.
Nghọta nka nka
Njikọ nfe ahụ na-atụgharị ọrụ ngọngọ ọ bụla ka ọ bụrụ y = F(x) + x. N'oge mgbasagharị azụ, gradient na-agafe na ụzọ mkpirisi njirimara agbanweghị, yabụ na ọ nweghị ike ịla n'iyi na nso efu ọbụlagodi n'ofe narị ọkwa. Nke a na-eme ka nchịkọta miri emi nwee ike ịzụ ya. Ụzọ mkpirisi njirimara agbakwunyeghị paramita ọzọ; naanị mgbe ntinye na nha mmepụta dị iche na-eme ka obere ntule (1x1 convolution) na-edozi akụkụ tupu mgbakwunye.
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 netwọkụ ndị fọdụrụnụ
Njikọ ndị fọdụrụ ugbu a dị nso na ụwa niile: ndị ntụgharị, ụdị mgbasa ozi, na ụdị asụsụ buru ibu niile na-eji ha eme ka ọzụzụ nke nchịkọta dị omimi guzosie ike. Nnyocha na-aga n'ihu na ụdị dị iche iche dị ka ResNets tupu arụ ọrụ, ụzọ ResNeXt agbakọtara, na ijikọta echiche ndị fọdụrụ na ọzụzụ na-enweghị nhazi. Na-atụ anya na isi ụkpụrụ mwụpụ-njikọ ga-adịgide dị ka ihe mgbochi ụlọ na-adịghị mma, ọbụnadị ka ihe owuwu ndị gbara ya gburugburu na-apụ na mgbagha dị ọcha na nlebara anya na ụdị ngwakọ.
Mmejuputa n'ezie n'ụwa
Ọkpụkpụ azụ nke nhazi nke ImageNet (ResNet-50, ResNet-101) ejiri mee ihe dị ka ndị na-ewepụta atụmatụ a zụrụ azụ maka ịnyefe mmụta.
Achọpụta etuto na ọnya na redio redio na ihe onyonyo site na iji ihe ngbanwe dabere na ResNet
Nchọpụta ihe na ihe atụ nkewa dịka ngwa ngwa R-CNN na Mask R-CNN na-eji ọkpụkpụ azụ ResNet.
Pipeline nghọta ịnya ụgbọ ala nke na-ekewa ndị na-agafe agafe, ụgbọ ala, na akara sitere na okpokolo agba igwefoto
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 Residual Networks 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
Njirimara netwọk pyramid
Ajụjụ a na-ajụkarị
What is Residual Networks?
Netwọk ndị fọdụrụnụ (ResNets) bụ netwọkụ akwara dị omimi nke na-agbakwunye 'njikọ sụọ' na-ekwe ka akwa akwa mụta obere mgbanwe kama mgbanwe zuru oke. Aghụghọ a dị mfe mere ka o kwe omume ịzụ netwọkụ ọtụtụ narị ọkwa miri emi, na-akpalite mmụba na njiri mara onyonyo.
Kedu nsogbu njikọ ndị fọdụrụ edoziziri kpọmkwem?
Tupu ResNets, ịgbakwunye ọkwa ndị ọzọ mere ka izi ezi mebie ọbụna na data ọzụzụ. Mwụpụ njikọ doziri nke a site n'ime ka oyi akwa dị mfe iji bulie ya.
Kedu ihe ngọngọ fọdụrụ na-agbakọ n'ezie dị ka mmepụta ya?
Mpụta ihe mgbochi fọdụrụnụ y = F(x) + x, na-agbakwunye ihe fọdụrụ amụtara na ntinye site na njikọ mwụpụ.
Kedu ihe kpatara ịmafe njikọ na-enyere gradients aka n'oge ọzụzụ?
Ụzọ mkpirisi njirimara ahụ na-enye ụzọ kpọmkwem maka gradients ka ọ na-aga azụ na-agbanweghị agbanwe, na-egbochi nsogbu na-apụ n'anya na mkpọkọ dị omimi.
Ihe fọrọ nke nta ka ọ bụrụ ọkwa ole ka ụdị ResNet na-emeri sitere na 2015 nwere?
ResNet-152, nke nwere ọkwa 152, meriri asọmpi ImageNet nke afọ 2015, na-egosi na a ga-azụzi netwọk dị omimi nke ọma ugbu a.
Ọ bụrụ n'ikuku ngọngọ fọdụrụ amụta F(x) = 0, gịnị ka ngọngọ ahụ na-eme?
Mgbe F(x) = 0, mmepụta bụ naanị x, yabụ ngọngọ na-aghọ nkewa njirimara. Nke a na-eme ka akwa akwa ndị ọzọ ghara ịdị njọ ma ọ bụrụ na ọ dịghị mkpa.