Segment lu nekk model
Modèlu Segment Anything (SAM) mooy Meta modelu fondaasioŋ AI ngir xaaj nataal: suñu ko joxee benn poñ, benn boyet, wala benn tegtal bu gàtt, ci saasi dafay wane mbir mi ko méngoo.
Résumé
It was built to generalize to objects and images it never saw during training, making segmentation a promptable task.
Plongeur bu xóot
Meta AI moo ko génne ci 2023, SAM dafay soppi xaaj bi muy jafe-jafe buñu mëna laaj: nga jox ko benn laaj (bës, benn boyet, benn mask, wala ab tegtal bu juge ci bind) mu delloo benn wala masku mbir yu bari. Dooleem mingi aju ci yaatuwaayam: dañu ko tàggat ci SA-1B, muy benn done bu ëpp 1 milyaar ci mask ci 11 milioŋ ciy nataal, ñu tabax ko ci benn motëru annotation model-in-the-loop. Ci wàllu architecture, SAM amna encodeur nataal bu diis buy daw benn yoon ci nataal bu nekk, encodeur prompt bu woyof, ak decodeur mask bu gaaw, suko defee benn nataal buñ samp mën nañu ko laajaat ci saa si. Dafay may toxal zero-shot ci liggéey yu bari. SAM 2, biñu génne ci 2024, dafay yokk lii ci wideo, di topp mbir ci biir kadre yi.
Gis-gis xarala
SAM dafay jëfandikoo enkodeer nataal buy soppi gis-gis (ViT), ñu koy faral di tàggat ak enkodeer automaske, ngir génne nataal bu dëgër. Laajte yi dañu leen kode ci ay jeton, ak dekodeer bu sukkandiko ci transformatër ak cross-attention fuses jetons yi ak nataal biñ samp ci mask yiy génne boole ci poñ wóolu. Ngir saafara jafe-jafe yi (bës mën na tekki butoŋu, chemise, wala nit), SAM dafay wax mask yu bari yu baax benn yoon ba noppi rang leen, bàyyi jëfandikoo downstream wala laaj yu gëna bari di dimbiguate.
njeextalu pexe
Gaawaay ak yaatuwaay
Visual IA mën na otomatise saytu, gis ak etiketu liggéey ci eskaal.
Tabax tànneef
Ekipu kreatif yi mën nañu defar konsept yu gëna gaaw te duñu def lu bari ci loxo.
Ekip ak def liggéey
Liggéeyukaay yi mën nañu jëfandikoo siñaal nataal wala wideo yu jafewoon lool ci liggéey.
Ëlëgu segment xeetu model
SAM nekk na yaxu ndigg biñ jagleel jumtukaayi annotation yi, nataali pajum, robotik, ak pipeline AR, ñu koy faral di boole ak detektër wala xeetu mbind ngir vocabulaire bu ubbeeku 'segment by name' workflows. Xaarandil anam yu gëna woyof, gëna gaaw (MobileSAM, EfficientSAM) ngir jëfandikoo ci aparey, boole bu gëna xóot ak làkk ngir segmentation bu lalu ci bind, ak wéyal yokkute ci wideo ak 3D. Bi mu nekkee xeetu fondaasioŋ, dañuy gëna jëfandikoowaat ay embeddings ci ni couche perception buy dundal yeneen sistem.
Doxal ci àdduna dëgg
Platform yiy joxe leeral ci nataal yi dañuy jëfandikoo SAM ngir may etiketkat yi ñu klike benn yoon ba noppi defar seen bopp ay masku mbir yu leer, di wàññi diiru etiket bi.
Gëstukat yi dañuy méngale SAM (lu melni, MedSAM) ngir wane cër yi ak tumër yi ci CT ak MRI scans.
Editëru nataal ak wideo yi dañu boole SAM ngir dagg sujet yi wala dindi lënd yi ci benn klik.
SAM 2 dafay topp ak xaaj mbir yi ci biir kadre wideo ngir am effet AR ak gis-gis robotik.
Risk yi ak balustrade yi
Yelleefi nataal ak nangu mën na nekk risku yoon sudee fi ñu bawoo leerul.
Performance model bi mën na wuute ci leeraay bi, demographie bi ak environmaa bi.
Njuumteg positive yi mën nañu dem te kenn duko seetlu fileek xool wuñu buntu wóolu sa bopp.
Roadmap ngir samp gi
Mandargal kritërium nangug njub, woowaat ak njëgu njuumte.
Saytu ak done yu méngoo ak anam yi ñuy liggéeyee dëgg.
Yokk jàngat nit ngir xam fu wóorul dara wala am njeexital yu rëy.
Toppal model drift bi nga baaxal ko ginaaw bi kamera bi wala done yi soppeekoo.
Weyal di banneexu
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 Segment Anything Model 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
Gis bi ci topp
Royuwaayi deggoo
Laaj yi ñuy faral di laaj
What is Segment Anything Model?
Modèlu Segment Anything (SAM) mooy Meta modelu fondaasioŋ AI ngir xaaj nataal: suñu ko joxee benn poñ, benn boyet, wala benn tegtal bu gàtt, ci saasi dafay wane mbir mi ko méngoo. Ñu tabaxoon ko ngir mu jëm ci mbir ak nataal yu mu musul gis ci diiru tàggat, tax segmentation nekk liggéey bu gaaw.
Ban xalaat moo tax SAM nekk 'modèle fondation' ci segmentation?
SAM dafay reframe segmentation ci nimu ko mënee, ñu defaree ko ngir toxal zero-shot ci mbir ak nataal yi nekk ci biti setu tàggat bi.
Lu tollu ci ban yaatuwaayu SA-1B lañuy jëfandikoo ngir tàggat SAM?
SA-1B amna lu ëppu benn miliyaar ciy mask ci kaw luy tollu ci 11 milioŋ ciy nataal, ñu tabax ko ci benn motëru annotaasioŋ bu nekk ci biir loop bi.
Lan moo tax SAM xaaj architecture bi ci encodeur nataal bu diis ak encodeur/decodeur bu woyof?
Encodage nataal bu seer bi dafay dox benn yoon; ginaaw ga encodage prompt bu yomb ak decodage mask may gaaw, weccoo xalaat laajaat benn nataal bi.
naka lay def ba SAM mëna jëflante ak laaj bu leerul, lu melni benn klik bu mëna tekki mbir yu bari?
SAM dafay genne ay masku kàndidaa yu bari ak poñ suko defee ñu mëna saafara jafe-jafe yi ci rang wala yeneen laaj.
Ban xeetu yaxu ndigg la SAM di jëfandikoo ngir enkode nataalu dugal bi?
Encodeur nataalu SAM Transformateur Vision la, ñu koy faral di tàggat ci autoencoding buñ maske, muy defar nataal bu dëgër.