Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
Unbalanced U-Net°ú GAN(Generative Adversarial Networks)À» ÀÌ¿ëÇÑ Çѱ¹¾î ÆùÆ® ÀÚµ¿ º¯È¯ |
¿µ¹®Á¦¸ñ(English Title) |
Automatic Transformation of Korean Fonts using Unbalanced U-net and Generative Adversarial Networks |
ÀúÀÚ(Author) |
¹æ°¡
°í½ÂÇö
¹æ¾ç
Á¶±Ù½Ä
Pangjia
Seunghyun Ko
Yang Fang
Geun-sik Jo
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 46 NO. 01 PP. 0015 ~ 0021 (2019. 01) |
Çѱ۳»¿ë (Korean Abstract) |
º» ³í¹®¿¡¼´Â ¿ø¹® ÆùÆ®¸¦ ƯÁ¤ÇÑ ¾Æ³¯·Î±× ÆùÆ® ½ºÅ¸ÀÏ·Î º¯È¯Çϴ ŸÀÌÆ÷±×·¡ÇÇ º¯È¯ ¹®Á¦ ¿¡ ´ëÇØ ¿¬±¸ÇÑ´Ù. ŸÀÌÆ÷±×·¡ÇÇ º¯È¯ ¹®Á¦¸¦ ÇØ°áÇϱâ À§ÇØ ÀÌ ¹®Á¦¸¦ À̹ÌÁö¿Í À̹ÌÁö ¹ø¿ª ¹®Á¦·Î Ä¡ ȯÇÏ°í GANÀ» ±â¹ÝÀ¸·Î ÇÑ ¾ð¹ë·±½º Çü u-net ¾ÆÅ°ÅØó¸¦ Á¦¾ÈÇÑ´Ù. ±âÁ¸ÀÇ ¹ë·±½º Çü u-net°ú´Â ´Þ¸® Á¦¾ÈÇÏ´Â ¾ÆÅ°ÅØó´Â ¾ð¹ë·±½º Çü u-netÀ» Æ÷ÇÔÇÑ µÎ °³ÀÇ ¼ºê³ÝÀ¸·Î ±¸¼ºµÈ´Ù. (1)¾ð¹ë·±½º Çü u-netÀº ÀÇ¹Ì ¹× ±¸Á¶ Á¤º¸¸¦ À¯ÁöÇÏ¸é¼ Æ¯Á¤ ±Û²Ã ½ºÅ¸ÀÏÀ» ´Ù¸¥ ½ºÅ¸ÀÏ·Î º¯È¯ÇÑ´Ù. (2) GANÀº L1 ¼Õ½Ç, »ó¼ö ¼Õ½Ç ¹× ¿øÇÏ´Â ¸ñÇ¥ ±Û²ÃÀ» »ý¼ºÇÏ´Â µ¥ µµ¿òÀÌ µÇ´Â ÀÌÁø GAN ¼Õ½ÇÀ» Æ÷ÇÔÇÏ´Â º¹ÇÕ ¼Õ½Ç ÇÔ¼ö¸¦ »ç ¿ëÇÑ´Ù. ½ÇÇè°á°ú Á¦¾ÈÇÏ´Â ¸ðµ¨ÀÎ ¾ð¹ë·±½º Çü u-netÀÌ ¹ë·±½º Çü u-net º¸´Ù cheat loss¿¡¼ ºü¸¥ ¼ö·Å ¼Óµµ¿Í ¾ÈÁ¤ÀûÀÎ Æ®·¹ÀÌ´× ¼Õ½ÇÀ» º¸¿´°í generate loss¿¡¼ Æ®·¹ÀÌ´× ¼Õ½ÇÀ» ¾ÈÁ¤ÀûÀ¸·Î ÁÙ¿©¼ ¸ðµ¨ ¼º ´É Ç϶ô ¹®Á¦¸¦ ÇØ°áÇÏ¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
t In this paper, we study the typography transfer problem: transferring a source font, to an analog font with a specified style. To solve the typography transfer problem, we treat the problem as an image-to-image translation problem, and propose an unbalanced u-net architecture based on Generative Adversarial Network(GAN). Unlike traditional balanced u-net architecture, architecture we proposed consists of two subnets: (1) an unbalanced u-net is responsible for transferring specified fonts style to another, while maintaining semantic and structure information; (2) an adversarial net. Our model uses a compound loss function that includes a L1 loss, a constant loss, and a binary GAN loss to facilitate generating desired target fonts. Experiments demonstrate that our proposed network leads to more stable training loss, with faster convergence speed in cheat loss, and avoids falling into a degradation problem in generating loss than balanced u-net.
|
Å°¿öµå(Keyword) |
À̹ÌÁö¿Í À̹ÌÁö ¹ø¿ª
ÆùÆ® º¯È¯
GAN ¼Õ½Ç
º¹ÇÕ ¼Õ½Ç
image to image translation
typography transfer
GAN loss
compound loss
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|