SakeTami
mariannasimnett
mariannasimnett

patreon


More progress pics

These show some test data sets from the shoot (it's still training - more on that later), sculptures in the making and some bts shots of me as Athena

no sharing please! x

More progress pics More progress pics More progress pics More progress pics More progress pics More progress pics More progress pics More progress pics

Comments

Thanks Elisabetta! yes of course, this is made using a model called styleGAN. Here's a link to explain how it works: https://www.analyticsvidhya.com/blog/2021/05/stylegan-explained-in-less-than-five-minutes/ It benefits from having similar footage (subject facing the same direction, locked off camera, etc) it then trains on the images and tries to rebuild them to a realistic level. The surprises and mistakes are the best part for me, like where my face turns into a bunch of grapes and so on. Other models I played around with were 'VQGAN' and 'diffusion models'. This all sounds like a headache... because it is! I am working with Arash Akbari (https://www.arashakbari.com/) who is proficient in custom AI. I'm fascinated by the way this open source technology is growing exponentially and I'm learning a lot, fast! :)

Info Simnett

Hi Marianna, this looks incredible! I look forward to seeing the outcome of the AI training. Would you mind sharing which AI platform you are using? I would be interested to know. Thanks!


More Creators