[mac] VToonify [...ing]
ยท
๐Ÿ‘พ Deep Learning
VToonify paper https://arxiv.org/abs/2209.11224 ์›๋ณธ ์ด๋ฏธ์ง€์˜ input ๊ฐ’์„ ์–ด๋Š ์ •๋„ degree๋กœ ๋ณ€ํ™˜ํ• ์ง€ ๋ณ€๊ฒฝ ๊ฐ€๋Šฅ ๊ธฐ์กด์˜ StyleGAN์—์„œ ์˜์ƒ์„ Toonify ์ง„ํ–‰์‹œ ๋ฐœ์ƒํ•˜๋Š” ๋‹จ์ ์ธ ํ•ด์ƒ๋„ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ–ˆ๋‹ค. Toonify์™€ DualStyleGAN ๋ชจ๋ธ ๋‘˜๋‹ค ์ง€์› (www.mmlab-ntu.com/project/vtoonify/) multi-scale content condition์œผ๋กœ ์›๋ณธ ์ด๋ฏธ์ง€์™€ stylegan์˜ ๋ณ€ํ™˜ ์ด๋ฏธ์ง€๋ฅผ ํ•ฉ์ณค๋‹ค.(Trainable Fusion Module ์‚ฌ์šฉ) ๊ธฐ์กด StyleGAN์€ ์˜์ƒ ๋ณ€ํ™˜์‹œ ๋ถ€์ž์—ฐ์Šค๋Ÿฌ์›€์„ ๊ฐœ์„  reference style์— ๋”ฐ๋ผ ๋ณ€ํ™˜ ๊ฐ€๋Šฅ pre-trained model ์ œ๊ณต (์ •๋ง ๊ฐ์‚ฌ!) ์ƒ์—…์  ์‚ฌ์šฉ ..
๋‹คํ–ˆ๋‹ค
'vtoonify' ํƒœ๊ทธ์˜ ๊ธ€ ๋ชฉ๋ก