Overview
GigaGAN ibhirioni-parameta GAN inoratidza generative adversarial network inogona kukwira kune zvinyorwa-kune-mufananidzo chizvarwa, ichikwikwidza mamodheru ekupararira uku ichigadzira mifananidzo mazana enguva nekukurumidza.
GigaGAN Yakaremerwa Jenareta ndeyekombuta-yekuona workflows inodudzira kana kuburitsa inooneka midhiya yekuongorora, mashandiro, uye kugadzira.
Deep Dive
GigaGAN, yakaunzwa neAdobe nevatsvaguri muna 2023, yakapokana nefungidziro yekuti maGAN aisakwanisa kukwira semhando dzekuparadzira. Pakutanga maGAN mahombe akadai seStyleGAN-XL akanetseka kudzidzisa zvakatsiga pamahombe, akasiyana dhata. GigaGAN yakagadzirisa izvi nekuwedzera jenareta uye kusarura, kuwedzera bhangi reyakadzidziswa convolution mafirita akasarudzwa pa-sample, uye nekubatanidza kutarisisa kune mameseji embeddings. Yakadzidziswa pamabhiriyoni emifananidzo-mavara maviri, 1-bhiriyoni-parameta jenareta inogadzira mufananidzo we512px mukati memasekonzi 0.13, inokurumidza kupfuura iyo inodzokorodza denoising yekupararira. Iyo zvakare inotsigira yakadzika-nzvimbo kududzira, kusanganisa maitiro, uye yakaparadzana GAN-yakavakirwa upsampler inogona kushandura 128px yekupinda kuita inopinza 4K mufananidzo.
Technical Insight
Iyo kiyi yehunyengeri ndeye 'sample-adaptive kernel sarudzo' module: pachinzvimbo cheimwe yakagadziriswa convolution seti, jenareta inobata bhanga remafirita uye inoshandisa iyo inomisikidzwa kuverengera uremu hunovasanganisa pamufananidzo. Zvakasanganiswa nekudzidziswa kwakawanda uye musarura anotonga zvigamba pane zvakati wandei zvigadziriso uye zvinoenderana nezvinyorwa zveCLIP, izvi zvinodzikamisa kudzidziswa kweanopikisa pachiyero apo maGAN akambodonha.
Kudzidzira GigaGAN Yakayerwa Jenareta
GigaGAN ibhirioni-parameta GAN inoratidza generative adversarial network inogona kukwira kune zvinyorwa-kune-mufananidzo chizvarwa, ichikwikwidza mamodheru ekupararira uku ichigadzira mifananidzo mazana enguva nekukurumidza. GigaGAN Yakaremerwa Jenareta ndeyekombuta-yekuona workflows inodudzira kana kuburitsa inooneka midhiya yekuongorora, mashandiro, uye kugadzira. Kuti uvake kunzwisisa kwakadzama, tora GigaGAN Scaled Generators semuenzaniso wekushanda, kwete chinhu chimwe chete: tsanangura zvinodikanwa, kujekesa fungidziro, uye kuparadzanisa izvo hurongwa hunogona kuita zvakavimbika kubva kune izvo zvichiri kuda kutonga kwenyanzvi.
Mukuita, zvikwata zvakasimba zvinoshandisa GigaGAN Scaled Generators kuenzanisa kurongeka nezvinoitika zvekushanda semhando yedata, kusiyana kwemwenje, uye kuenderana kwemazita. Ivo vanonyora zvakajeka maitiro ebudiriro, bvunzo vachipokana ne data rechokwadi uye mafambiro ebasa, uye iterate zvichibva pane zvakacherechedzwa maitiro ekutadza kwete kuhwina-nguva imwe chete yebhenji. Apa ndipo apo kunzwisisa kwe theoretical kunoshanduka kuve kugona kwakasimba pane chigadzirwa, mutemo, uye mashandiro.
Visual AI inogona kuita otomatiki yekuongorora, yekuona, uye yekumaka mabasa pachiyero. Panguva imwecheteyo, kodzero dzeMufananidzo uye kubvumirwa kunogona kuve njodzi dzepamutemo kana hunhu husina kujeka. Nzira yakatsiga ndeyekubatanidza kukurumidza kuyedza nekutonga: mhanyisa vatyairi vendege, tora humbowo, buritsa matanda esarudzo, uye urambe uchivandudza chengetedzo semaitiro emuenzaniso, zvinotarisirwa nemushandisi, uye zvinodikanwa zvekutonga.
Strategic Impact
Visual AI inogona kuita otomatiki yekuongorora, yekuona, uye yekumaka mabasa pachiyero.
Visual AI inogona kuita otomatiki yekuongorora, yekuona, uye yekumaka mabasa pachiyero. Mukutumirwa kwemhando yepamusoro, izvi zvinoshandurirwa kuita mitemo inoyerwa yekushanda, miganhu yevaridzi, uye tsika dzekudzokorora dzinodzokororwa kuitira kuti zvikwata zvikwire kuvimba pane kukwidza kusajeka.
Zvikwata zvekugadzira zvinogona prototype pfungwa nekukurumidza nekudzokororwa kwemaoko mashoma.
Zvikwata zvekugadzira zvinogona prototype pfungwa nekukurumidza nekudzokororwa kwemaoko mashoma. Mukutumirwa kwemhando yepamusoro, izvi zvinoshandurirwa kuita mitemo inoyerwa yekushanda, miganhu yevaridzi, uye tsika dzekudzokorora dzinodzokororwa kuitira kuti zvikwata zvikwire kuvimba pane kukwidza kusajeka.
Mashandisirwo anogona kushandisa masaini emifananidzo nemavhidhiyo ayo aimbove akaoma kugadzirisa.
Mashandisirwo anogona kushandisa masaini emifananidzo nemavhidhiyo ayo aimbove akaoma kugadzirisa. Mukutumirwa kwemhando yepamusoro, izvi zvinoshandurirwa kuita mitemo inoyerwa yekushanda, miganhu yevaridzi, uye tsika dzekudzokorora dzinodzokororwa kuitira kuti zvikwata zvikwire kuvimba pane kukwidza kusajeka.
Real-World Implementation
Kugadzira mufananidzo we 512px kubva pane zvinyorwa zvinokurumidza mukati mechikamu chegumi chesekondi yekudyidzana dhizaini yekutarisisa.
Kukwidza yakaderera-resolution 128px foto kune yakanatswa 4K mufananidzo uchishandisa iyo GAN-based super-resolution upsampler.
Kupindirana zvinyoro-nyoro pakati pekukurudzira kuviri munzvimbo yakanyarara kuti ifambise shanduko, sekapu yekofi inopinda mu teapot.
Kushandisa masitayera kusanganisa kuchengetedza chimiro chechinyorwa uku uchichinjanisa chimiro chayo chehunyanzvi kana pendi yeruvara muAdobe-maitiro ekugadzirisa maturusi.
Maitiro Ekuita
GigaGAN Yakasimudzwa Jenareta mukuita
Kugadzira mufananidzo we512px kubva pamashoko ekuchimbidza mukati mechikamu chimwe chete chegumi chesekondi yekudyidzana dhizaini yekutarisisa.
Kugadzira mufananidzo we512px kubva pamashoko ekuchimbidza mukati mechikamu chegumi chesekondi yekudyidzana dhizaini yekutarisisa Matimu anowanzo kuwana mibairo iri nani kana achinge atsanangura hunhu hwepamberi, chengetedza nzira yekukwira kwevanhu yemakesi ekumucheto, uye kuteedzera zvese zvakawanikwa zvechigadzirwa uye mutengo wekukanganisa nekufamba kwenguva.
GigaGAN Yakasimudzwa Jenareta mukuita
Kukwidza yakaderera-resolution 128px pikicha kune yakanatswa 4K mufananidzo uchishandisa iyo GAN-based super-resolution upsampler.
Kukwidza yakaderera-resolution 128px pikicha kune yakajeka 4K chifananidzo uchishandisa iyo GAN-yakavakirwa super-resolution upsampler Matimu anowanzo kuwana mhedzisiro iri nani kana achinge atsanangura emhando yepamusoro kumberi, chengetedza nzira yekukwira kwevanhu yemakesi emupendero, uye kuteedzera zvese zvakawanikwa zvechigadzirwa uye mutengo wekukanganisa nekufamba kwenguva.
GigaGAN Yakasimudzwa Jenareta mukuita
Kupindirana zvinyoro-nyoro pakati pezvinokurudzira zviviri munzvimbo yakadzikama yekumhanyisa shanduko, sekapu yekofi ichipinda mu teapot.
Kupindirana zvinyoro-nyoro pakati pekukurudzira kuviri munzvimbo yakadzikama yekumhanyisa shanduko, senge kapu yekofi ichipinda mu teapot Matimu anowanzo kuwana mibairo iri nani kana achinge atsanangura mhando yepamusoro kumberi, chengetedza nzira yekukwira kwevanhu yemakesi emupendero, uye kuteedzera zvese zvakawanikwa zvechigadzirwa nemitengo yekukanganisa nekufamba kwenguva.
GigaGAN Yakasimudzwa Jenareta mukuita
Kushandisa kusanganiswa kwemaitiro kuchengetedza chimiro chechidzidzo uku uchichinjanisa maitiro ehunyanzvi kana pendi yeruvara muAdobe-maitiro ekugadzirisa maturusi.
Kushandisa dhizaini kusanganisa kuchengetedza chimiro chechinyorwa uku uchichinjanisa chimiro chayo chehunyanzvi kana pendi yeruvara muAdobe-maitiro ekugadzirisa maturusi Matimu anowanzo kuwana mhedzisiro iri nani kana achinge atsanangura hunhu hwepamberi, chengetedza nzira yekukwira kwevanhu yemakesi ekumucheto, uye kuteedzera zvese zvakawanikwa zvechigadzirwa uye kukanganisa mutengo nekufamba kwenguva.
Njodzi & Guardrails
Kodzero dzemifananidzo uye kubvumirwa kunogona kuve njodzi dzepamutemo kana provenance isina kujeka.
Kuita kwemuenzaniso kunogona kusiyanisa kupenya, huwandu hwevanhu, uye nharaunda.
Manyepo enhema anogona kusacherechedzwa kunze kwekunge zvikumbaridzo zvekuvimba zvikatariswa.
Implementation Roadmap
Tsanangura maitiro ekugamuchirwa echokwadi, kurangarira, uye mutengo wekukanganisa.
Tsanangura maitiro ekugamuchirwa echokwadi, kurangarira, uye mutengo wekukanganisa. Bata nhanho yega yega segedhi rehumbowo: kana maitiro asina kusangana, imbomira kuburitsa, vhara gaka, uye wobva wawedzera kushandiswa.
Edzai nedata rinoenderana nemamiriro chaiwo ekugadzira.
Edzai nedata rinoenderana nemamiriro chaiwo ekugadzira. Bata nhanho yega yega segedhi rehumbowo: kana maitiro asina kusangana, imbomira kuburitsa, vhara gaka, uye wobva wawedzera kushandiswa.
Wedzera ongororo yemunhu kune yakaderera-kusavimbika kana yakakwirira-inokanganisa kufanotaura.
Wedzera ongororo yemunhu kune yakaderera-kusavimbika kana yakakwirira-inokanganisa kufanotaura. Bata nhanho yega yega segedhi rehumbowo: kana maitiro asina kusangana, imbomira kuburitsa, vhara gaka, uye wobva wawedzera kushandiswa.
Tevera modhi kudonha uye simbisa mushure mekuchinja kwekamera kana dataset.
Tevera modhi kudonha uye simbisa mushure mekuchinja kwekamera kana dataset. Bata nhanho yega yega segedhi rehumbowo: kana maitiro asina kusangana, imbomira kuburitsa, vhara gaka, uye wobva wawedzera kushandiswa.