AMABWIRIZA

LAION na Gufungura imibare

LAION numudage udaharanira inyungu wasohoye amashusho manini afunguye-yamakuru yamakuru, azwi cyane LAION-5B, yongereye imbaraga mumahugurwa yuburyo bubyara umusaruro nka Stable Diffusion.

2 min somaIbiherutse kuvugururwa

Incamake

It matters because it made web-scale multimodal data freely available to researchers outside large corporations.

Kwibira cyane

LAION (Urwego runini rwa Artificial Intelligence Open Network) ni umudage udaharanira inyungu washinzwe mu 2021 kugirango uhindure demokarasi ubushakashatsi bwo kwiga imashini urekura imibare minini ifunguye. Isohora rizwi cyane, LAION-5B, ikubiyemo hafi miliyari 5.85 z'amashusho-inyandiko ebyiri zungururwa zivuye mu mbuga rusange za Crawl ukoresheje __AIU_PROTECTED_10 __ ya CLIP yerekana kugirango ugumane ibice aho ibisobanuro n'amashusho bihuye. Icy'ingenzi, LAION ntabwo yakira amashusho ubwayo; ikwirakwiza URL na metadata, abakoresha rero bakuramo amashusho kuva kurubuga rwumwimerere. Izi mibare zagize uruhare runini mu mahugurwa ahamye Diffusion hamwe nizindi fungura inyandiko-ku-shusho. LAION yahuye n’igenzura rikomeye: mu 2023 abashakashatsi basanze isano iri hagati y’amashusho y’ihohoterwa mu buryo butemewe muri dataset, bituma LAION ayimanura, ayisukura, kandi yongera gusohora verisiyo itekanye, agaragaza ingaruka ziterwa no gusiba ku rubuga rwa interineti.

Ubushishozi

LAION-5B yubatswe mugusikana Crawl Rusange ya HTML yerekana amashusho ya HTML hamwe na alt-text, hanyuma ukoresheje CLIP kugirango ubare isano iri hagati ya buri shusho ninyandiko zayo. Byombi munsi yikigereranyo cya cosine bisa nkaho byajugunywe, hasigaye gusa guhuza amashusho-inyandiko byombi byasigaye. Dataset igabanijwe nururimi kandi ikubiyemo mbere yo kubara CLIP yashyizwemo, ituma gushakisha byihuse. Kuberako URL gusa zibitswe, ihuza kubora buhoro buhoro itesha agaciro imyororokere mugihe.

Ingaruka z'Ingamba

Vendor strategy

Ibishushanyo mbonera byabacuruzi bigira ingaruka kubiranga ikipe yawe ishobora kubaka ubutaha.

Igiciro na bije

Amagambo yubucuruzi nuburyo bwo kohereza bigira ingaruka kubiciro byigihe kirekire ningaruka.

Risk and safety

Isosiyete ishimangira gushiraho ibicuruzwa bitemewe, igihagararo cyumutekano, no gufungura.

Kazoza ka LAION no gufungura imibare

Gufungura imibare myinshi yimibare izahura nigitutu cyiyongera kubijyanye nuburenganzira, uburenganzira, nibintu byangiza, biganisha ku kuyungurura gukomeye, gukusanya uruhushya-byemewe, no guhitamo kwiyandikisha. LAION yongeye kurekura dataset isukuye byerekana impinduka zijyanye no kugenzura umutekano nkintambwe isanzwe. Tegereza amakuru yubukorikori cyangwa yemewe, ibipimo ngenderwaho, hamwe nibikoresho byo gutahura. Ubushyamirane hagati yo gufungura laboratoire ntoya hamwe ningaruka zemewe n’imyitwarire y’amakuru yakuwe ku rubuga bizasobanura icyiciro gikurikira cyo kubaka dataset.

Gushyira mu bikorwa Isi

Amahugurwa afunguye inyandiko-y-ishusho yerekana urugero nka Diffusion ihamye kuri miliyari ya shusho-yanditswemo

Kubaka no gupima ibipimo bya CLIP-ishusho-inyandiko yo kugarura na sisitemu yo gutondekanya zeru

Ubushakashatsi kuri dataset kubogama, umutekano wibirimo, hamwe no kwerekana amakuru kurwego rwurubuga

Kurungurura ibice ukoresheje ururimi, gukemura, cyangwa amanota meza kugirango ukore imibare yihariye

Ingaruka & Kurinda

Gutangiza amatangazo arashobora gusumbya ituze mubikorwa nyabyo byakazi.

Ibiciro bya API cyangwa guhindura politiki birashobora guhagarika ibitekerezo ijoro ryose.

Abashoramari bonyine bishingira byongera gufunga no kwimuka.

Igishushanyo mbonera

1

Suzuma abatanga serivisi ukoresheje imirimo yawe bwite na datasets.

2

Ongera usuzume ubuzima bwite, umutekano, namategeko mbere yo kwishyira hamwe.

3

Komeza gahunda yo gusubira inyuma kurugero cyangwa abacuruzi.

4

Kurikirana inyandiko zisohora kugirango impinduka zumuhanda ntizitangaje amakipe.

Komeza Ubushakashatsi

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 LAION and Open Datasets quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Tangira ikibazo

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Ubuyobozi bukurikira

Gufungura isoko AI

Ibibazo bikunze kubazwa

What is LAION and Open Datasets?

LAION numudage udaharanira inyungu wasohoye amashusho manini afunguye-yamakuru yamakuru, azwi cyane LAION-5B, yongereye imbaraga mumahugurwa yuburyo bubyara umusaruro nka Stable Diffusion. Nibyingenzi kuko byatumye urubuga rwa interineti rwimibare myinshi rushobora kuboneka kubushakashatsi hanze yamasosiyete manini.

Niki LAION ikwirakwiza cyane cyane mumashusho-yamakuru yamakuru?

LAION ntabwo yakira amashusho; ikwirakwiza URL na metadata, bityo abakoresha bazana amashusho kuva kurubuga rwabo rwambere.

Hafi ya bangahe ishusho-inyandiko ebyiri ziri muri LAION-5B?

LAION-5B ikubiyemo amashusho agera kuri miliyari 5.85, ni yo mpamvu '5B' mu izina ryayo.

Ni ubuhe buryo LAION yakoresheje mu kuyungurura amashusho-inyandiko zombi kugirango zihuze ubuziranenge?

LAION yakoresheje __AIU_PROTECTED_10 __ ya CLIP kugirango apime amashusho-yanditswemo asa kandi ajugunye byombi bidahuye.

Ni ubuhe buryo bukomeye bwo gufungura ibintu bwahuguwe hakoreshejwe amakuru ya LAION?

Stable Diffusion, ifungura inyandiko-y-ishusho yerekana, yatojwe kuri LAION ishusho-yamakuru.

Ni ikihe kibazo gikomeye abashakashatsi bavumbuye muri LAION-5B muri 2023?

Abashakashatsi basanze amasano yibikoresho byo guhohotera abana bitemewe, bituma LAION imanura dataset hasi, kuyisukura, no kongera gusohora verisiyo itekanye.