Awọn ile-iṣẹ Itọsọna

LAION ati Ṣii Datasets

LAION jẹ ai-jere kan ti ara ilu Jamani ti o ṣe idasilẹ awọn ipilẹ data ọrọ-iṣiro aworan ti o tobi pupọ, olokiki julọ LAION-5B, eyiti o mu ikẹkọ ti awọn awoṣe ipilẹṣẹ ṣiṣi bi Itupalẹ Stable.

Gbọ si itọsọna yiiAI-ṣe afihan · ~2 min

Lori iwe yi2 min ka
  1. Akopọ
  2. Jin Dive
  3. Real-World imuse
  4. Ojo iwaju ti LAION ati Ṣii Datasets
  5. Fifi o sinu iwa
  6. Tesiwaju Ṣiṣawari
  7. Awọn ibeere ti a beere nigbagbogbo

Akopọ

O ṣe pataki nitori pe o jẹ ki data multimodal iwọn wẹẹbu larọwọto wa si awọn oniwadi ni ita awọn ile-iṣẹ nla.

Jin Dive

LAION (Large-scale Artificial Intelligence Open Network) is a German nonprofit founded in 2021 to democratize machine learning research by releasing large open datasets. Its best-known release, LAION-5B, contains roughly 5.85 billion image-text pairs filtered from Common Crawl web data using OpenAI's CLIP model to keep pairs where the caption and image align. LAION distributes URLs and metadata while the original image files remain on their web hosts. The dataset helped researchers train and study open image-text models, including Stable Diffusion. In December 2023, researchers reported URLs to suspected child sexual abuse material in LAION-5B, and LAION withdrew the dataset for a safety revision. In August 2024, LAION described Re-LAION versions created by comparing existing records with partner-provided hash lists. The release excludes known suspect links identified by those partners as of July 2024; LAION explicitly states that filters cannot guarantee all harmful content has been removed from web-scale data.

Real-World imuse

Ikẹkọ ṣiṣi ọrọ-si-aworan awọn awoṣe bii Itupalẹ Stable lori awọn ọkẹ àìmọye ti awọn orisii aworan

Ilé àti àṣeyẹ̀wò ìṣàtúnṣe àwòrán ara CLIP ara-ìsọ̀rọ̀ àti àwọn ọ̀nà ìsọ̀rọ̀-ìsọt-odo

Ṣiṣayẹwo aiṣedeede dataset, ailewu akoonu, ati iṣafihan data ni iwọn wẹẹbu

Sisẹ awọn ipin nipasẹ ede, ipinnu, tabi Dimegilio darapupo lati ṣẹda awọn ipilẹ data ti n ṣatunṣe amọja

Ojo iwaju ti LAION ati Ṣii Datasets

Ṣiṣii awọn iwe data multimodal yoo dojukọ titẹ ti ndagba ni ayika aṣẹ-lori-ara, ifọkansi, ati akoonu ipalara, titari si sisẹ sisẹ ti o lagbara, gbigba-nimọ iwe-aṣẹ, ati jade awọn iforukọsilẹ. Atun-itusilẹ LAION ti ipilẹ data mimọ kan ṣe ifihan agbara iyipada kan si iṣayẹwo ailewu bi igbesẹ aiyipada kan. Reti diẹ sii sintetiki tabi data ti a fun ni iwe-aṣẹ, awọn iṣedede ododo, ati ohun elo wiwa. Aifokanbale laarin iraye si ṣiṣi fun awọn ile-iṣẹ kekere ati ofin ati awọn ewu iṣe ti data oju opo wẹẹbu yoo ṣalaye ipele atẹle ti kikọ data.

Fifi o sinu iwa

Kini idi ti o ṣe pataki

  • Ilana olutaja. Awọn maapu opopona olutaja ni ipa kini awọn ẹya ti ẹgbẹ rẹ le kọ ni atẹle.
  • Iye owo ati isuna. Awọn ofin iṣowo ati awọn aṣayan imuṣiṣẹ ni ipa lori idiyele igba pipẹ ati eewu.
  • Ewu ati ailewu. Awọn imoriya ile-iṣẹ ṣe apẹrẹ awọn abawọn ọja, iduro ailewu, ati ṣiṣi.

Ṣọra fun

  • Awọn ikede ifilọlẹ le ju iduroṣinṣin lọ ni awọn iṣan-iṣẹ iṣelọpọ gidi.
  • Ifowoleri API tabi awọn iyipada eto imulo le fọ awọn arosinu ni alẹ.
  • Igbẹkẹle olutaja ẹyọkan ṣe alekun titiipa-inu ati awọn idiyele ijira.

Next awọn igbesẹ

  1. Ṣe ayẹwo awọn olupese nipa lilo awọn iṣẹ ṣiṣe tirẹ ati awọn ipilẹ data.
  2. Ṣe atunyẹwo asiri, aabo, ati awọn ofin ofin ṣaaju iṣọpọ.
  3. Ṣetọju eto ipadabọ kọja awọn awoṣe tabi awọn olutaja.
  4. Bojuto awọn akọsilẹ itusilẹ nitoribẹẹ awọn iyipada maapu oju-ọna ma ṣe iyalẹnu awọn ẹgbẹ.

Tesiwaju Ṣiṣawari

Ìwé ìròyìn ọfẹ

Gba ìkànìyàn AI ojoojúmọ́

Ìtàn AI mẹta tó jẹ́rìí ní gbogbo owurọ́ ọjọ́ iṣẹ́, kọ́wé ní Gẹ̀ẹ́sì tó rọrùn. Ọfẹ́ fún gbogbo àkókò, kò sí ìpolówó.

Ìmeèlì kan ní gbogbo ọjọ́ iṣẹ́. Yọkúrò ní ìkànìyàn ní ìkànìyàn kan. A kò ta tàbí pín àdírẹ́sì rẹ.

Ṣe idanwo fun ara rẹ

Take the LAION and Open Datasets quiz

Ìdáhùn lẹ́sẹ̀kẹsẹ̀ lórí gbogbo ìdáhùn, àti iwe-ẹ̀rí tó lè pin pẹ̀lú ID tó lè jẹ́rìí lẹ́yìn tí o bá kọ́ ẹ̀kọ́.

Bẹrẹ adanwo

Ṣe àtìlẹ́yìn ẹ̀kọ́ AI ọfẹ́. AI Understanding jẹ́ àjọ àìní èrè 501(c)(3) — kò sí ìpolówó, kò sí àkúnya owó, ní gbogbo àkókò. Ṣe ẹbun kan

Awọn ibeere ti a beere nigbagbogbo

Kini LAION ati Ṣii Datasets?

LAION jẹ ai-jere kan ti ara ilu Jamani ti o ṣe idasilẹ awọn ipilẹ data ọrọ-iṣiro aworan ti o tobi pupọ, olokiki julọ LAION-5B, eyiti o mu ikẹkọ ti awọn awoṣe ipilẹṣẹ ṣiṣi bi Itupalẹ Stable. O ṣe pataki nitori pe o jẹ ki data multimodal iwọn wẹẹbu larọwọto wa si awọn oniwadi ni ita awọn ile-iṣẹ nla.

Kini LAION ṣe pinpin ni akọkọ ninu awọn iwe data ọrọ-aworan rẹ?

LAION ko gbalejo awọn aworan; o pin awọn URL ati metadata, nitorinaa awọn olumulo mu awọn aworan lati awọn orisun oju opo wẹẹbu atilẹba wọn.

O fẹrẹ to awọn orisii ọrọ-aworan melo ni o wa ni LAION-5B?

LAION-5B ni nipa 5.85 bilionu awọn orisii ọrọ aworan, nitorinaa '5B' ni orukọ rẹ.

Awoṣe wo ni LAION lo lati ṣe àlẹmọ awọn orisii ọrọ-aworan fun didara titete?

LAION lo CLIP OpenAI lati ṣe iwọn ibajọra aworan ati sọ awọn orisii ibaamu ti ko dara.

Awoṣe ipilẹṣẹ ṣiṣi olokiki wo ni o jẹ ikẹkọ ni lilo data LAION?

Itankale Stable, awoṣe ọrọ-si-aworan ti o ṣii, ni ikẹkọ lori data ọrọ-aworan LAION.

What issue led LAION to withdraw LAION-5B in December 2023?

Researchers identified URLs to suspected CSAM in LAION-5B. LAION withdrew the dataset and later described a revision using partner-provided hash lists; this does not establish that every suspicious link was live or that every possible harmful item has been removed.