Tshuab kev kawm tshuab rau cov duab synthesis thiab txo suab nrov hauv cov duab hmo ntuj

Stability AI tau luam tawm cov qauv npaj rau Stable Diffusion tshuab kev kawm, muaj peev xwm ntawm kev sib txuas thiab hloov kho cov duab raws li cov lus piav qhia hauv cov lus ntuj. Cov qauv muaj ntawv tso cai raws li kev tso cai Creative ML OpenRAIL-M daim ntawv tso cai rau kev lag luam siv. Los cob qhia cov kab ke, pawg ntawm 4000 NVIDIA A100 Ezra-1 GPUs thiab LAION-5B sau, suav nrog 5.85 billion dluab nrog cov lus piav qhia, tau siv. Yav dhau los, cov cai rau cov cuab yeej rau kev cob qhia neural network thiab tsim cov duab tau qhib-qhov chaw raws li daim ntawv tso cai MIT.

Qhov muaj ntawm cov qauv npaj tau tsim thiab tsim qauv tsim nyog uas tso cai rau ib qho pib sim ntawm PC nrog tus qauv GPUs tau coj mus rau qhov tshwm sim ntawm ntau qhov haujlwm cuam tshuam:

  • textual-inversion (code) - ib qho ntxiv-on uas tso cai rau koj los tsim cov duab nrog tus cwj pwm, yam khoom lossis style. Nyob rau hauv thawj Stable Diffusion, cov khoom nyob rau hauv cov duab synthesized yog random thiab uncontrolled. Cov ntawv thov ntxiv tso cai rau koj ntxiv koj tus kheej cov khoom pom, khi rau cov ntsiab lus thiab siv lawv hauv kev sib txuas.

    Piv txwv li, hauv Stable Diffusion tsis tu ncua koj tuaj yeem hais kom lub kaw lus tsim cov duab nrog "miv hauv lub nkoj". Tsis tas li ntawd, koj tuaj yeem qhia meej txog cov yam ntxwv ntawm miv thiab lub nkoj, tab sis nws yog qhov tsis paub tseeb uas cov miv thiab lub nkoj yuav ua ke. Textual-inversion tso cai rau koj los cob qhia cov kab ke ntawm cov duab ntawm koj cov miv lossis lub nkoj thiab ua ke cov duab nrog ib tus miv lossis lub nkoj. Ib qho zoo sib xws, nws tuaj yeem hloov cov duab nrog qee yam khoom, ua piv txwv ntawm qhov pom kev zoo nkauj rau kev sib txuas, thiab qhia cov ntsiab lus (piv txwv li, los ntawm tag nrho cov kws kho mob, koj tuaj yeem siv cov kev xaiv zoo dua thiab zoo. hauv qhov xav tau style).

    Tshuab kev kawm tshuab rau cov duab synthesis thiab txo suab nrov hauv cov duab hmo ntuj

  • ruaj khov-diffusion-animation - tsim cov duab animated (tsav) raws li kev sib cuam tshuam ntawm cov duab tsim hauv Stable Diffusion.
  • stability_diffusion.openvino (code) - qhov chaw nres nkoj ntawm Stable Diffusion, uas tsuas yog siv CPU rau kev suav, uas tso cai rau kev sim ntawm cov tshuab uas tsis muaj zog GPUs. Yuav tsum muaj lub processor txhawb nqa hauv OpenVINO lub tsev qiv ntawv. Raws li txoj cai, OpenVINO muab cov plugins rau Intel processors nrog AVX2, AVX-512, AVX512_BF16 thiab SSE txuas ntxiv, ntxiv rau Raspberry Pi 4 Model B, Apple Mac mini thiab NVIDIA Jetson Nano boards. Unofficially, nws muaj peev xwm siv OpenVINO ntawm AMD Ryzen processors.
  • sdamd yog qhov chaw nres nkoj rau AMD GPUs.
  • Thawj qhov kev siv ntawm video synthesis.
  • ruaj khov-diffusion-gui, ruaj khov-diffusion-ui, Artbreeder Collage, diffuse-the-so - graphical interfaces rau tsim cov duab siv Stable Diffusion.
  • beta.dreamstudio.ai, Hugging Face Spaces, hlky Stable Diffusion WebUI - web interfaces rau duab synthesis siv Stable Diffusion.
  • Plugins rau kev koom ua ke Stable Diffusion nrog GIMP, Figma, Blender thiab Photoshop.

Tsis tas li ntawd, peb tuaj yeem nco ntsoov cov ntawv tshaj tawm los ntawm Google ntawm cov cai ntawm RawNeRF (RAW Neural Radiance Fields) tshuab kev kawm, uas tso cai, raws li cov ntaub ntawv los ntawm ntau cov duab RAW, txhawm rau txhim kho qhov zoo ntawm cov duab nrov nrov hauv qhov tsaus ntuj thiab hauv. teeb pom kev tsis zoo. Ntxiv nrog rau kev tshem tawm cov suab nrov, cov cuab yeej tsim los ntawm qhov project ua kom muaj kev nthuav dav, tshem tawm glare, ua ke HDR thiab hloov tag nrho cov teeb pom kev zoo hauv cov duab, nrog rau rov tsim cov duab peb sab ntawm cov khoom siv ob peb daim duab los ntawm ntau lub kaum ntse ntse, hloov lub ntsiab lus ntawm kev pom, tswj kom pom tseeb thiab tsim cov duab txav.

Tshuab kev kawm tshuab rau cov duab synthesis thiab txo suab nrov hauv cov duab hmo ntuj
Tshuab kev kawm tshuab rau cov duab synthesis thiab txo suab nrov hauv cov duab hmo ntuj


Tau qhov twg los: opennet.ru

Ntxiv ib saib