Xamme Jëf
Xamme jëf mooy jàngal ordinatër yi ñu xàmmee li nit ñi wala mbir yi *def* ci wideo — daw, yëngu, daanu, ubbi buntu — du li feeñ ci benn kadre kese.
Résumé
It matters because understanding motion over time unlocks applications from sports analytics to elderly fall detection.
Plongeur bu xóot
Xamme jëf demna lu ëpp xaaj nataal bu taxawul ci xalaat ci ni pixel yi di soppeekoo ci diir bi. Benn kadre mën na wane nit ku nekk ci jawwu ji; li ñuy toppalante rek mooy wane ndax dañuy tëb, dañuy daanu wala ñuy féey. Sistem yu njëkk ya dañu defaree loxo ay man-mani yëngu yu melni flow optik ak trajectoire yu dëgër. Xeetu xam-xam bu bees bi dafay jëfandikoo reso yu xóot yi: architecture yu am ñaari yoon ñuy jëfandikoo melokaan (kadre RGB) ak yëngu (flux optique) ñu tàqaloo; Reseau convolutionnel 3D (lu melni C3D ak I3D) dañuy jaar ci bayaal *ak* waxtu; ak soppi wideo (TimeSformer, VideoMAE) dañuy bàyyi xel ci wàllu barab ak jamono. Benchmark yiñ miin bokkuna ci Kinetics (700 njàngale jëf nit ci YouTube), UCF101, ak Dara-Dara, luy forse model yi ñu xam yoon wi ci jamono moo gën muy wax ci seen.
Gis-gis xarala
Jafe-jafe bi gëna mag mooy modele dimension temporel bi. Convolution 3D dafay yokk filtre 2D bu baax ak axis bu xóot buy jaar ci kadre yu bari, suko defee mu jàng motif yiy yëngu ci saasi. Kaf I3D dafay 'inflate' diisaay yi ci reso nataal 2D biñ njëkka tàggat ci ImageNet ci 3D ci di leen toppandoo ci diir bi, di joxe point de départ bu dëgër. Ñaari-flux pexe lu moy feed precomputed optique flow ci bànxaas bu wuute, encoding leer mouvement ba noppi fusing ko ak melokaan màndarga.
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 Gëstu Jëf
Doxalin bi mingi toxu ci transformatëri wideo yu baax ak tàggat sa bopp (modeling wideo bu maske) buy jàngee ci wideo yu amul etiket, dagg koolute ci annotation yu seer. Xaarandi lëkkaloo bu gëna dëgër ak xeeti làkk yu bari suko defee sistem yi baña yam ci joxe etiket ci jëf yi waaye ñu mëna leen leeral ak xalaat ci làkk wiñ nàmp. Saa yu dëggu, xàmmee ci aparey yi ngir wearables, robotik, ak kamera yu xarañ, frontiere bu mag la, ci wetu xàmmee bu baax biy wuutale yëngu yu woyof, yu nuru luy nuru.
Doxal ci àdduna dëgg
Sistem yiy gis daanu ci këri mag ñi di artu liggéeykat yi su nit ki daanu, di wuutale daanu ak toog wala tëdd
Platform jàngat sport yiy etiketu ci saasi ay serwiis, tacle ak tiire ci wideo match ngir anterneer ak di siiwal mbir yu am solo
Wottu ak saytu kaaraange giy wane jeffin ju jaarul yoon lu melni xeex, wëréelu, wala kuy yéeg ci ñag
Interface yuñ kontre ak yëngu-yëngu ak aplikaasioŋu fitness yuy lim reps yi ak xool formu tàggat yaram ci xàmmee tànki yaram yi ci diir bu gàtt
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 Action Recognition 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
Xamme kanam
Laaj yi ñuy faral di laaj
What is Action Recognition?
Xamme jëf mooy jàngal ordinatër yi ñu xàmmee li nit ñi wala mbir yi *def* ci wideo — daw, yëngu, daanu, ubbi buntu — du li feeñ ci benn kadre kese. Dafa am solo ndax xam yëngu ci diir bi dafay ubbi aplikaasioŋ yi tàmbalee ci jàngat tàggat yaram ba ci gis mag ñi daanu.
Lan mooy wuutale xàmmee jëf ak xaaj nataal buñ miin?
Modèlu xàmmee jëf dafay yëngu ci kaw ay kadre yu toppalante, ndax benn nataal bu taxaw mënul wane ndax nit mingi tëb, daanu wala féey.
Ci reso buy xàmmee jëf bu am ñaari stream, lan la ñaareelu stream bi di faral di def?
Architecture yu am ñaari stream dañuy jëfandikoo benn bànxaas ngir melokaan (kadre RGB) ak ñaareelu bànxaas ngir flow optik, biy kode bu leer dem bi ak dikk bi ci diggante kadre yi.
Lan la convolusioŋ 3D di yokk suñu ko méngale ak convolusioŋ 2D buñ miin?
Benn convolution 3D dafay yokk filtre bi ci dimension xóotaayu/waxtu, may ko mu jàng motif yiy yëngu ci kadre yu toppalante.
Lan mooy 'inflation' bi xeetu I3D bi di jëfandikoo?
I3D dafay 'inflate' diisaay yiñ njëkka tàggat ci nataali 2D (lu melni ImageNet) ci 3D ci di leen kopie ci axis temporal bi, di joxe ab ndoorte bu dëgër.
Lan moo waral setu done Something-Something siiw ci xàmmee jëf?
Something-Something dañu ko tëral ci anam wuy tax jëf ji a ngi aju ci doxalinu jamono ak yëngu, di tere model yi njuuj njaaj ci jëfandikoo fond wala melokaanu mbir kese.