Aventures in neuronale rendering : bi jagokɛlaw bɛ ka baarakɛcogo ɲɛnamaw sɛmɛntiya cogo min na

Neural rendering bɛ ka fɛn caman sɛmɛntiya jagokɛlaw ka kunnafoni yelenw dilancogo, u ɲɛnabɔcogo, ani u bilacogo la, u kɛtɔ ka kalan jugu ni ɔridinatɛri laadala ja pipelines fara ɲɔgɔn kan. K’a ta 3D jɔli kokura waati yɛrɛ la ka se AI ka fɛn dilannenw ja ma, ɲɛnajɛ minnu bɛ kɛ ni hakilijagabɔ ye, olu bɛ nafa yelenw di ɲɔgɔndan na jɛkuluw ma minnu b’a fɛ ka nin dancɛ in sɛgɛsɛgɛ.

Neural Rendering ye mun ye tigitigi ani mun na jago ɲɛmɔgɔw ka kan k'u janto u la ?

Sɛgɛsɛgɛli sigilen bɛ masin kalanni ni ɔridinatɛri ja cɛtigɛyɔrɔ la, a bɛ baara kɛ ni hakilijagabɔlanw ye walasa ka fototalanw, 3D ɲɛjiraliw, ani sigida fangamaw fara ɲɔgɔn kan ka bɔ kunnafonidilanw na. A tɛ i n’a fɔ laadala rendering pipelines minnu bɛ u jigi da bololabaarakɛlaw ka shading models kan, neural approaches bɛ kalan kɛ k’a ɲɛsin diɲɛ lakika misaliw ma — ka bɔli kɛ minnu tun tɛ se ka kɛ fɔlɔ walima minnu sɔngɔ tun ka gɛlɛn kosɛbɛ ka bɔ.

Jagokɛlaw ɲɛmɔgɔw fɛ , o nafa ka bon bawo ɲɛjirali kɔnɔkow tɛ daɲɛw labɛncogo ɲuman ye tugun . O ye bonyalan ye min bɛ se ka suman. Jagokɛyɔrɔ minnu bɛ ɛntɛrinɛti kan, minnu bɛ baara kɛ ni AI ka fɛn dilannenw ye, olu bɛ katalogu dilanni sɛgɛsɛgɛli teliyalenw fɔ. Jagokɛlaw ka jɛkulu minnu bɛ baara kɛ ni neural style transfer ye, olu bɛ dɔ bɔ dilancogo iterasiyɔn waati la. Architecte ni fɛn dilannikɛlaw prototype teliya la ni neural 3D reconstructions ye ka bɔ foto nɔgɔmanw na.

ye

"Jɛkulu minnu bɛna se sɔrɔ san tan nataw la, olu tɛ wajibi ye minnu ka daɲɛgafew ka jɛkuluba bɛ yen — olu de bɛ hakilitigiya kɛ ka AI-driven visual production (AI-driven visual production) baara kɛ ni hakɛ ye."

Fɛɛrɛ jɔnjɔn jumɛnw bɛ senfagabana sɛgɛsɛgɛli fɛɛrɛ kɔfɛ ?

Masin minnu bɛ senfagabana jirali kɔfɛ, olu faamuyali bɛ jagokɛlaw dɛmɛ u ka se ka denfaya latigɛ hakilitigiya la. Foro in jɔlen bɛ fɛɛrɛ damadɔw kan minnu bɛ ɲɔgɔn dafa, u kelen-kelen bɛɛ bɛ bɛn baarakɛcogo danfaralenw ma:

• Neural Radiance Fields (NeRF): NeRF modɛliw bɛ 3D ɲɛjiraliw jira i n’a fɔ volumɛtiri baarakɛcogo tɛmɛnenw minnu bɛ kalan ja 2D fɛ. Ja damadɔ bɛ se ka fototalan 360 degɛrɛ lakikaw dilan — nafa ka bon kosɛbɛ fɛn dilannenw ja la, taama virtuwɛliw, ani filaninw nizɛrikan na.
• Jɛkulu juguw (GANw) : GANw fanga bɛ ja sinsinni , cogoya cili , ani sɛbɛnniw bɔli la . Jagokɛlaw bɛ baara kɛ n’u ye fɛn bɛɛ la k’a ta AI ka avatarw la ka se sosiyete ka kunnafonidilanw kɔnɔkow ma minnu bɛ kɛ otomatiki la sɛgɛsɛgɛli la.
• Differentiable Rendering : Nin fɛɛrɛ in b’a to gradients ka woyo rendering pipelines fɛ, o bɛ kɛ sababu ye ka sistɛmuw dɛmɛ u ka 3D geometry kalan k’a ɲɛsin 2D kɔlɔsili ma — o nafa ka bon kosɛbɛ augmented reality applications ani product configurators fɛ.
• Deferred Neural Rendering : Ni jateminɛ ni yecogo faralen don ɲɔgɔn kan, fɛɛrɛ minnu bɛ bila kɔfɛ, olu bɛ sira di yeelen, fɛnw ani filɛlifɛnw sɛgɛsɛgɛli ma waati yɛrɛ la — ka da wuli ka ɲɛsin feereli kɛcogo jɛlenw ma ani jiralikɛnɛw ma.
• Diffusion Models : Neural rendering bɔnsɔn laban bɛ baara kɛ ni diffusion-based architectures ye walasa ka yecogo gɛlɛnw dilan ani k’u ladilan ni kɔrɔsili ye min ɲɔgɔn ma deli ka kɛ fɔlɔ, sisan u bɛ sɔrɔ kosɛbɛ jago APIw fɛ.

Diɲɛ tɔnba minnu bɛ diɲɛ kɔnɔ , olu bɛ ka hakilijagabɔ kɛ cogo di baarakɛcogo la ?

danfara min bɛ ɲininiw jirali ni fɛn dilanni bilali cɛ , o dɔgɔyara kosɛbɛ . Jɛkulu minnu bɛ seko ni dɔnko siratigɛ la, olu bɛ ka hakilijagabɔ kɛcogo don baarakɛcogo kɔrɔw la ni jateminɛw ye minnu bɛ se ka suman.

Fɛɛrɛfeerelaw bɛ ka fɛn sɔngɔ gɛlɛnw fototalanw bila NeRF-based renders nɔ na minnu bɛ bɔ smartfon ka ja la. SKU kelen min tun bɛ tile tilancɛ studio kalan de wajibiya fɔlɔ, sisan o bɛ se ka kɛ kɛrɛ tan ni caman na ani yeelen cogoyaw la lɛrɛ damadɔ kɔnɔ. Mode taamasiyɛnw bɛ baara kɛ ni GANw ye walasa ka waati kɔnɔ lookbook ja dilan, A/B bɛ yecogo hakilinaw kɔrɔbɔ ka sɔrɔ k’u yɛrɛ bila farikoloɲɛnajɛ la.

Architecture ni real estate firms bɛ danfaralanw bila senkan walasa ka duguma plan kunnafoniw sɛmɛntiya ka kɛ 3D taamacogo ye min bɛ ɲɔgɔn sɔrɔ. Kunnafonidila minnu bɛ u sen don kosɛbɛ fotoko ɲɛjiraliw la minnu bɛ kɛ ni fotoko ye, o bɛ feereli siraw surunya. Fɛn dilanni na, pibiliki filanin nimɔrɔma minnu bɛ baara kɛ ni hakilijagabɔ ye, olu b’a to jogoɲini jɛkuluw bɛ se ka ɲɛnayeli nafamafɛnw dɔnni ladege sani farikolo prototypes ka sɔrɔ.

Fɛn min bɛ tali kɛ waleyali ɲɛtaa la, o ye jɛ-ka-baara ye : hakili-la-ko-minɛn minnu bɛ tali kɛ kɔnɔkow ɲɛnabɔli la, porozɛw ɲɛnabɔli la, ani jagokɛyɔrɔw la, olu bɛ nafa caman di. AI baarakɛminɛn minnu bɛ u yɛrɛ la, olu bɛ ɲɔgɔn sɔsɔ; AI min bɛ don a kɔnɔ, o bɛ flow da.

Sɛgɛsɛgɛli minnu bɛ tali kɛ ɲɔgɔn na, olu bɛ mun jira hakilijagabɔ la ni laadala fɛɛrɛw ye ?

Laadalata rendering pipelines — minnu jɔlen bɛ farikoloɲɛnajɛ motɛriw kan i n’a fɔ Arnold, V-Ray, walima Cycles — bɛ se ka fɔ ka ɲɛ ani ka seko ni dɔnko kunnafoni finmanw di. U ye sanu sariya ye filimu ɲɛnama VFX ni fɛn dilanni ɲɛjirali sanfɛtaw la. Nka, u bɛ jatebɔ waatiba de wajibiya, fɛɛrɛko dɔnnikɛla ŋanaw, ani iterasiyɔn sɛrɛkili janw.

Neural rendering bɛ o deterministic control dɔw jago teliya, otomatiki ani ladamuni kama. Jagokɛlaw ka baarakɛcogo fanba la — jago ja, fɛn dilannenw yecogo, kalan kunnafoniw bɔli, jɛɲɔgɔnya ko kɛlenw — sisan hakililata fɛɛrɛw bɛ bɛn walima ka tɛmɛ laadala jogoɲumanya jateminɛw kan, musaka ni waati tilayɔrɔba dɔ la.

Kiritigɛ waleyali : hakilijagabɔ tɛ laadala pibilikiw nɔ bila baarakɛcogo bɛɛ la, nka a bɛ fɛn min bɛ se ka kɛ sɔrɔko siratigɛ la sugu cɛmancɛ jɛkuluw bolo, o bonya kosɛbɛ. Seko minnu tun maralen bɛ fɔlɔ baarakɛda minnu bolo minnu ka daɲɛgafe-sɛnɛyɔrɔbaw bɛ yen, sisan olu bɛ se ka sɔrɔ ekipu nɔgɔlenw fɛ minnu ka plateforme stack bɛnnen don.

Jagokɛlaw ka baarakɛminɛnw bɛ se ka jɛkuluw dɛmɛ cogo di ka hakilijagabɔ baarakɛcogo ɲɛnabɔ ?

Ka senfagabana sɛgɛsɛgɛli fɛɛrɛ ta, o ye kɛlɛ tilancɛ dɔrɔn ye. Porozɛw, jɛɲɔgɔnw, nafolo ani kiliyanw ka kumaɲɔgɔnya ɲɛnabɔli min bɛ yecogo lamini na, o bɛ baarakɛcogo sabatilenw de wajibiya. O yɔrɔ de la jagokɛcogo jɛlen bɛ kɛ fɛn nafama ye.

Mewayz — jagokɛla ka OS min bɛ se ka kɛ modulu 207 ye, baarakɛla 138.000 ni kɔ dalen bɛ min na — o bɛ baarakɛcogo kɔkanna di, daɲɛw ni fɛɛrɛbɔ jɛkuluw mago bɛ min na walasa ka hakilijagabɔ baarakɛcogo sɛgɛsɛgɛ k’a sɔrɔ marako ɲagami ma kɛ. Ni labɛnw bɛ daminɛ dɔrɔmɛ 19 dɔrɔn na kalo o kalo, Mewayz bɛ porozɛw ɲɛnabɔli, jɛkuluw ka jɛkafɔ, kiliyanw ka dakunw, CRM, faturaw, ani kɔnɔkow baarakɛcogo fara ɲɔgɔn kan ka kɛ kɛnɛ kelen ye. Ekipuw bɛ se ka render baara sinsinniw ɲɛnabɔ, ka kiliyanw ka jaabiw sɛrɛkiliw lajɛ, ka baarakɛlaw ka wari sarataw ɲɛnabɔ, ani ka poroze nafa kɔlɔsi k’a sɔrɔ u ma ɲɔgɔn falen-falen baarakɛminɛnw ni ɲɔgɔn cɛ minnu tigɛlen don. Ni hakilijagabɔw dilanni hakɛ bɛ ka bonya, Mewayz bɛ sɛgɛsɛgɛli kɛ kɛrɛfɛ — k’a lajɛ ko baarakɛcogo gɛlɛya tɛ kɛ buteli ye abada daɲɛw ni fɛɛrɛbɔ bɔli la.

Ɲininkali minnu bɛ kɛ tuma caman na

yala hakilijagabɔ bɛ tali kɛ baarakɛdabaw dɔrɔn de la minnu ka AI ekipuw y' u yɛrɛ di wa ?

A tɛ kɛ fewu . Demokarasi kɛli ni hakilijagabɔ ye sankaba kɔnɔ APIw fɛ ani baarakɛminɛn minnu bɛ se ka baara kɛ ni u ye, o y’a to o seko ninnu kɛra sababu ye ka se ka sɔrɔ jagokɛla misɛnninw fɛ, dabɔbaa yɛrɛmahɔrɔnw fɛ, ani sugu cɛmancɛ tɔnw fɛ. Neural rendering baarakɛminɛn caman tɛ kodɔn dɔnniya si wajibiya, wa u bɛ don ɲɔgɔn na k’u ɲɛsin dilancogo ni jago baarakɛcogo kɔrɔw ma. O kun ye ka baarakɛminɛn ɲumanw fara ɲɔgɔn kan ni baarakɛyɔrɔ ye min bɛ porozɛw labɛn ani ka ekipuw bila ɲɔgɔn na.

A ka c'a la, a bɛ waati joli ta ka ROI ye ka bɔ hakilijagabɔ tali la ?

ROI waatibolodalenw bɛ danfara ka kɛɲɛ ni baarakɛcogo ye, nka jɛkulu minnu bɛ laadala fɛn dilannenw fotota baarakɛcogo bila u nɔ na, olu ka teli ka nafa ɲumanw ye sɛnɛko fɔlɔ kɔnɔ — tuma dɔw la dɔgɔkun damadɔ kɔnɔ. Tɔnɔba minnu bɛ sɔrɔ kosɛbɛ, olu bɛ bɔ studio musakaw dɔgɔyali la, waati teliya ka taa sugu la nafolo yelenw na, ani se ka kɔnɔkow caman ɲɔgɔnna caman dilan A/B sɛgɛsɛgɛli kama k’a sɔrɔ dɔ ma fara baarakɛnafolo kan ka kɛɲɛ ni ɲɔgɔn ye. Baarakɛcogo ɲuman bɛ dɔ fara waati kɔnɔ ni jɛkuluw bɛ kɔnɔna dɔnniyaw yiriwa ani k’u ka baarakɛcogo nɔgɔya.

jagokɛlaw ka kan ka baarakɛ gɛlɛya jumɛnw makɔnɔ ni u bɛ hakilijagabɔ kɛcogo ta ?

Baara gɛlɛya minnu ka ca kosɛbɛ, olu ye nafolomafɛnw dosiyebaw ɲɛnabɔli ye ekipu tilalenw kɔnɔ, ka hakilinaw labɛncogo ɲɛnabɔ fɛɛrɛko dɔnnikɛlaw ni fɛɛrɛtigiw cɛ minnu tɛ fɛɛrɛtigi ye, ka porozɛ musakaw nɔfɛtaama render baara caman kɔnɔ, ani ka taamasiyɛnw ka ɲɔgɔndɛmɛ sabati AI ka sɔrɔw kɔnɔ. O gɛlɛyaw tɛ dan ni hakilijagabɔ ye — u ye porozɛ ni jɛkulu ɲɛnabɔli gɛlɛya jɔnjɔnw ye, jago OS jɛlen i n’a fɔ Mewayz dabɔra kɛrɛnkɛrɛnnenya la ka ɲɛnabɔ.

Aventures in neuronale rendering tɛ yɔrɔjan kɔlɔsiliw ye minnu dan ye ɲininikɛyɔrɔw dɔrɔn ye — u ye sababu ɲɛnamaw ye, minnu labɛnnen don ka baara kɛ, minnu bɛ sɔrɔ jɛkuluw bolo minnu bonya bɛ se ka kɛ bi. Jagokɛlaw minnu bɛ taa sisan, olu bɛna nafa caman sɔrɔ ɲɔgɔndan na, minnu bɛ kɛ yecogo teliya la, musaka dɔgɔyali la, ani daɲɛw dilanni seko la.

I labɛnnen don ka baarakɛcogo jusigilan jɔ min b’a to i ka jɛkulu ka tugu cogoya kura dɔw kɔ i n’a fɔ hakilijagabɔ kɛli k’a sɔrɔ i ma bɔnɛ porozɛw, baarakɛnafolo, walima kiliyanw ni ɲɔgɔn cɛ jɛɲɔgɔnyaw la wa? I ka Mewayz taama daminɛ bi app.mewayz.com — modulu 207, plateforme kelen, min bɛ daminɛ $19/kalo.