GUIDE IA visuel

xóotaayu lépp xóotaayu monokulaar

DepthAnything xeetu fondation la buy xayma fu pixel bu nekk sori ci benn nataal bu amul benn aparey buñ jagleel.

2 simili jàngDañu mujjee yeesal

Résumé

It made robust, general-purpose depth sensing cheap and accessible for anything from phones to robots.

Plongeur bu xóot

DepthAnything (2024, bi gëstukat yi génne, lu ci melni ñooñu ci TikTok / ByteDance ak HKU) dafay wax ci xayma xóotaayu monokulaar: di wax luy waaja am ci benn nataalu RGB. Li gëna am solo ci liggéey bi mooy: ludul yéem ci done yu néew yiñ am, ekip bi dafa defar benn motër buy etikete ci boppam lu tollu ci 62 milioŋ ciy nataal yuñ etiketewul, ñu jëfandikoo xeetu jàngalekat, ba noppi ñu tàggat benn ndongo ci corpus bu rëy bii. Loolu dafay joxe zero-shot bu dëgër ci biir kër yi, ci biti, ak ci seen yu wuute ak yeneen yi. Li njëkk mooy génne xóotaayu relatif bi (pixel yi gëna jege wala gëna sori, du meetar yu dëggu yi). DepthAnything V2 (diggu 2024) dafa gëna ñaw ay detay yu rafet ci tàggat jàngalekat bi ci done yu synthetik ak dëgg gu mat sëkk, ba noppi distil ci nataal yu dëggu, defar boor yu fësul ak njuumti mbir yu leer.

Gis-gis xarala

Dafay jëfandikoo enkodeer buy soppi gis-gis bu DINOv2 buy dundal boppu wax luy waaja am bu dëgër bu nuroo ak DPT. Li gëna am solo mooy distillation semi-supervisée: jàngalekat buñu tàggat ci done yuñ etikete dafay pseudo-etiketu ay milioŋ ciy nataal yuñ etiketewul, te ndongo dafay jàngee ci ñoom ñaar. V2 dafay weccoo etiketu dëgg yu am bruit ak done yu synthetik ak xóotaayu pixel-perfect, ba noppi distillewaat ci nataal yu dëggu, di moytu ñàkkum ak bruit bu xóotaayu annotation dëgg, boole ci tëye ay pexe yu fëgër.

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 xóotaayu lépp lu xóotaayu monokulaar

Xaarandi lëkkaloo bu gëna dëgër ci weer AR, kamera yu xarañ, ak robotik fu LiDAR biñ jagleel seer lool wala gëna rëy. Xeetu metrik yiy génne metre dëgg, boole ci model wideo yu am xóotaayu stabil ci diir bu gàtt (amul flicker ci diggante kadre yi), ñu ngi gëna dem bu baax. Ginaaw model yooyu dañuy wàññeeku ngir mëna dox ci aparey yi ci jamono dëgg, gis-gis 3D bu benn kamera dina nekk mënin buñ jagleel, di dundal ordinatër spatial, navigation autonome, ak tabaxaat seen 3D.

Doxal ci àdduna dëgg

Defar kàrtu xóotaayu ngir dawal flër bu dëggu (bokeh) ci nataali nataali telefon yu xarañ yu am benn lenti.

Joxe gis-gis 3D ci gaañ-gaañu ngir drone ak robot yu yomb te amul LiDAR wala kamera stéréo.

Sosal kàrtu kondisioneer xóotaayu ControlNet suko defee defarkati nataal yi baña yàq geometri seen.

Soppi nataal ak filmu 2D ci effet 3D wala parallax ngir VR ak feem stéréoskopik.

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

1

Mandargal kritërium nangug njub, woowaat ak njëgu njuumte.

2

Saytu ak done yu méngoo ak anam yi ñuy liggéeyee dëgg.

3

Yokk jàngat nit ngir xam fu wóorul dara wala am njeexital yu rëy.

4

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 DepthAnything Monocular Depth quiz

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

Tambalil quiz

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

Gis bi ci topp

Xayma xóotaayu monokulaar

Laaj yi ñuy faral di laaj

What is DepthAnything Monocular Depth?

DepthAnything xeetu fondation la buy xayma fu pixel bu nekk sori ci benn nataal bu amul benn aparey buñ jagleel. Defar na lu dëgër, lu yomb te mën nañu ko jëfandikoo ci lu nekk, liko dalee ci telefon wala ci robot.

Luy xayma xóotaayu 'monokular'?

Monokulaar mooy xóotaayu benn (mono) nataalu kamera, amul benn stéréo wala kaptër buy dox.

Lan moo nekkoon pexem DepthAnything ngir mëna am généralisation bu dëgër?

Ekip bi dafa defar benn motëru done buy pseudo-label fukki milioŋ ciy nataal yuñ amul etiket, muy yokk bu baax yaatuwaayu tàggat yaram.

Ban enkodeer la DepthAnything tabax ci kaw?

DepthAnything dafay jëfandikoo encodeur DINOv2 ViT boole ci boppu DPT bu dëgër buy wax luy am.

Ban xeetu xóotaayu DepthAnything mooy njëkka génne?

Modèle de base bi dafay wax luy waaja am ci xóotaayu relatif bi moo gën distance metric absolu bi.

naka la ko DepthAnything V2 gënee yombal ay detay yu rafet ak ay boor?

V2 tàggat jàngalekat bi ci nataali synthetik yu am xóotaayu ndànk, ba noppi distiye ko ci nataal yu dëggu ngir am ay fësal yu gëna ñaw.