Mgbagwoju anya na mgbagwoju anya
Mgbagwoju anya gbajiri agbaji (nke a na-akpọkwa convolutions atrous) na-etinye oghere n'etiti nha nzacha ka kernel wee kpuchie ebe buru ibu na-agbakwunyeghị paramita.
Nchịkọta
They let networks see wide context, crucial for segmentation and audio, while keeping resolution intact.
Ime miri emi
Ọkụ convolution nkịtị na-emetụ pikselụ dị n'akụkụ. Mgbasa ozi gbasaara na-agbasa otu nha kernel iche site na ọnụ ọgụgụ dilation, na-amapụ pikselụ n'etiti, yabụ kernel 3 × 3 nwere dilation 2 na-agbasa mpaghara 5 × 5 ka ọ na-eji naanị nha 9. Nke a na-agbasawanye ogige nnabata nke ukwuu mgbe ị na-eji ọnụma na-abawanye ụba, na-ahapụ netwọkụ ahụ ka ọ chịkọta ọnọdụ buru ibu na-enweghị ikpokọta ma ọ bụ steepụ nke ga-ebelata maapụ atụmatụ ahụ. Okwu atrous sitere na French a trous, nke pụtara nwere oghere. Nke a bara uru na nnukwu amụma ọrụ dị ka ngalaba semantic, ebe ịchọrọ ma nlele sara mbara yana mmepụta pikselụ ziri ezi, yana na WaveNet maka ịmegharị dabere na ọdịyo ogologo.
Nghọta nka nka
Ikpokọta convolutions gbasasịrị na ọnụego 1, 2, 4, 8 na-eto ubi nnabata dị ka ike nke abụọ ebe ọnụọgụ paramita na-adịgide. Atrous Spatial Pyramid Pooling (ASPP) na DeepLab na-agba ọsọ dị iche iche dilation ọnụego n'otu n'otu na fuses ha, na-ese ihe na otutu akpịrịkpa na otu ngafe. Ọnụego efu nwere ike ịkpata arịa ihicha, yabụ a na-ahọrọ ọnụego nke ọma ka mkpuchi kpuchie ya.
Mmetụta atụmatụ
Ọnụ ego na mmefu ego
Mkpebi ihe owuwu ụlọ na-akwalite arụmọrụ yana ọnụ ahịa ọrụ ruo ọtụtụ afọ.
Mkpebi doro anya
Nkà mmụta nka na-enyere ndị otu egwuregwu aka ịhọrọ nchịkọta ziri ezi, ọ bụghị naanị nke kachasị ọhụrụ.
Quality akara
Nhọrọ injinia ka mma na-ebelata ihe omume ntụkwasị obi na mmepụta.
Ọdịnihu nke Mgbagha na Mgbagwoju anya
Mgbanwe ndị gbawara agbawa na-anọkwa n'etiti ngalaba semantic na panoptic, onyonyo ahụike na ọgbọ ọdịyo. A na-ejikọta ha na nlebara anya, ebe dilation na-enye ohere nnabata ogologo oge dị ọnụ ala nke na-akwado nlebara anya onwe onye. Nchọpụta na-aga n'ihu na ọnụego mgbasa ozi na-agbanwe agbanwe na nke a ga-amụta na ịzere ngwa ngwa ngwa ngwa. Na-atụ anya ha n'ụdị usoro ogologo ogologo nke ọma yana nghọta ọnọdụ ọnọdụ maka sistemu kwụụrụ onwe ya.
Mmejuputa n'ezie n'ụwa
DeepLab na-eji mgbagwoju anya na ASPP maka ngalaba ntụgharị uche ọgbara ọhụrụ nke ihe nkiri okporo ụzọ.
WaveNet na-achịkọta mgbagha ihe kpatara iji mepụta ọdịyo na okwu dị adị
Nkewa onyonyo ahụike, dị ka etuto ma ọ bụ oke akụkụ ahụ, ebe okirikiri sara mbara yana nkọwa zuru oke ma ihe abụọ ahụ
Ntụle ọnọdụ n'ezie maka nghọta ịnya ụgbọ ala nke chọrọ nnukwu ogige nnabata na-enweghị atụfu mkpebi
Ihe ize ndụ & okporo ụzọ nche
Ịkwalite otu akara ngosi nwere ike zoo adịghị ike sistemụ sara mbara.
A na-eledakarị ihe akụrụngwa na ụgwọ ọrụ anya.
Ọdịiche nchekwa na nleba anya nwere ike itolite ka sistemu na-adịwanye mgbagwoju anya.
Map mmejuputa
Kọwaa latency, ịdịmma na ebumnuche ọnụ ahịa tupu mmejuputa ya.
Benchmark n'okpuru ibu dị adị na ọnọdụ data.
Nleba anya akụrụngwa maka mperi, ịkpafu na mmetụta onye ọrụ.
Kwadebe ụzọ nzaghachi azụghachi azụ na ihe omume tupu ịchachaa.
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 Dilated and Atrous Convolutions 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
Mgbagha ndị nwere ike kewapụta nke omimi
Ajụjụ a na-ajụkarị
What is Dilated and Atrous Convolutions?
Mgbagwoju anya gbajiri agbaji (nke a na-akpọkwa convolutions atrous) na-etinye oghere n'etiti nha nzacha ka kernel wee kpuchie ebe buru ibu na-agbakwunyeghị paramita. Ha na-ahapụ netwọk ka ịhụ ọnọdụ gbara gburugburu, dị oke mkpa maka nkewa na ọdịyo, ka ha na-edobe mkpebi na-emebibeghị.
Kedu ihe ọnụ ọgụgụ mgbasa ozi karịrị 1 na-eme kernel convolution?
Dilation oghere ndị dị arọ iche, na-amụba ogige nnabata na-abawanyeghị ọnụ ọgụgụ nke paramita.
Mkpụrụ ndụ 3x3 nke nwere ọnụ ọgụgụ mgbasa nke 2 na-agbasa nke ọma ka mpaghara buru ibu?
Site na dilation 2 enwere otu ọnọdụ wụsara n'etiti ịdị arọ nke ọ bụla, yabụ kernel na-agbatị mpaghara 5x5 ebe ọ ka na-eji naanị 9 arọ.
Kedu ihe kpatara convolutions gbasasịrị ji baa uru karịsịa maka nkewa semantic?
Nkewa chọrọ ọnọdụ sara mbara yana nkọwa ọkwa pikselụ n'otu oge; dilation na-eto gburugburu na-enweghị mbelata nke ikpokọta ma ọ bụ ịzọ ụkwụ ga-ebute.
Kedu ihe okwu atrous si abịa?
Atrous sitere na French uwe ogologo ọkpa, nwere oghere, na-akọwa oghere ndị etinyere n'etiti oke kernel.
Na WaveNet, gịnị kpatara convolutions gbasasịrị ekpokọtara bara uru?
Ọnụego dilation na-abawanye nke ukwuu na-ahapụ ihe ndabere WaveNet n'ofe puku kwuru puku nlele ọdịyo na-enweghị nnukwu kernels.