• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ³í¹®Áö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ³í¹®Áö > Çѱ¹Á¤º¸°úÇÐȸ ³í¹®Áö > Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) GANs ChainingÀ» ÀÌ¿ëÇÑ ÈÞ¸Õ ¾×¼Ç »ý¼º
¿µ¹®Á¦¸ñ(English Title) Pose ¬³ontrolled Creation of Human Actions Using GANs Chaining
ÀúÀÚ(Author) ¾ÆÁöÁî ½Ã¾ß¿¡ÇÁ   Á¶±Ù½Ä   Aziz Siyaev   Geun-Sik Jo  
¿ø¹®¼ö·Ïó(Citation) VOL 26 NO. 08 PP. 0390 ~ 0395 (2020. 08)
Çѱ۳»¿ë
(Korean Abstract)
Generative Adversarial Networks(ÀÌÇÏ GANs)Àº ´Ù¾çÇÑ ÄÜÅÙÃ÷ Á¦ÀÛ ºÐ¾ß¿¡¼­ ºü¸¥ ¹ßÀü ¼Óµµ¸¦ º¸¿©¿Ô´Ù. ƯÈ÷ ¿µ»ó »ý¼º ºÐ¾ß¿¡¼­´Â ¾Æ¹ÙŸ ¾Ö´Ï¸ÞÀ̼ǰú °°Àº Àΰ£ Á᫐ ÀÀ¿ë ÇÁ·Î±×·¥ÀÇ °³¹ß·Î Å« ÁÖ¸ñÀ» ¹Þ¾Ò´Ù. º» ³í¹®¿¡¼­´Â 2´Ü°è GAN ÆÄÀÌÇÁ¶óÀÎÀ» »ç¿ëÇÏ¿© ¡®GANÀ» ÅëÇÑ ¼øÂ÷Àû Àΰ£ Çൿ »ý¼º¡¯(Sequential Human Action Generation with GANs)¡¯¿¡ ´ëÇÑ ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. Á¦¾ÈÇÏ´Â ¹æ¹ýÀº ºñµð¿À »ý¼ºÀ» À§ÇØ ÀÚ¼¼ Á¦¾î °¡´ÉÇÑ Á¢±Ù ¹æ½ÄÀ» Àû¿ëÇÏ¿´´Ù. ÀÚ¼¼ »ý¼º±â·Î ÀÚ¼¼ÀÇ »À´ë¸¦ »ý¼ºÇÑ ÈÄ, ÇÁ·¹ÀÓ »ý¼º±â¸¦ ÅëÇØ ÅؽºÃ³¸¦ ÀÔÈ÷´Â ¹æ½ÄÀÌ »ç¿ëµÇ¾ú´Ù. ºñµð¿À Á¦ÀÛ¿¡ ´ëÇÑ ±¤¹üÀ§ÇÑ ½ÇÁ¦ ºÐ¼® °á°ú´Â Æ÷Áî Á¦¾î °¡´ÉÇÑ ºñµð¿À »ý¼ºÀÌ ¾ÈÁ¤ÀûÀÌ°í Á÷°üÀûÀ̶ó´Â °á°ú¸¦ º¸¿©ÁØ´Ù. °á°úÀûÀ¸·Î ´Ù¾çÇÑ ¿µ»ó Ç°Áú ½ÇÇèÀ» ÅëÇØ SeHAGANÀÇ ¼øÂ÷Àû Çൿ ¹æ¹ýÀÌ »ç¶÷ÀÇ ¿òÁ÷ÀÓ¿¡ ´ëÇÑ ÀÚ¿¬½º·¯¿î °íÇ°ÁúÀÇ ¿µ»óÀ» »ý¼ºÇÒ ¼ö ÀÖÀ½À» ÀÔÁõÇÏ¿´´Ù
¿µ¹®³»¿ë
(English Abstract)
The Generative Adversarial Networks (GANs) have shown rapid development in different content-creation tasks. Among them, the video generation has received attention because of the development of various human-centric applications such as avatar animation. The movie industry arranges vast research in automatic scenes generation and trailer production, which underlines the need for developing video generation technologies. In this paper, we propose our method for human actions video generation called SeHAGAN: Sequential Human Action Generation with GANs, which is a two-stage GANs pipeline. We apply the pose controllable approach for video generation. First, we produce the pose skeleton with our poses generator, and then we texture them with a frame generator. Extensive practical analysis in video production show that pose controllable video generation demonstrate stable and intuitive results. Various video quality experiments were conducted, and results illustrate that SeHAGAN generates a plausible, high-quality and realistic video of human movements.
Å°¿öµå(Keyword) ÈÞ¸Õ ¾×¼Ç »ý¼º   Æ÷Áî »ý¼º   ºñµð¿À »ý¼º   GAN   human actions generation   poses generation   video generation   GAN  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå