• 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¸Ã³¸®ÇÐȸ ³í¹®Áö > Á¤º¸Ã³¸®ÇÐȸ ³í¹®Áö ÄÄÇ»ÅÍ ¹× Åë½Å½Ã½ºÅÛ

Á¤º¸Ã³¸®ÇÐȸ ³í¹®Áö ÄÄÇ»ÅÍ ¹× Åë½Å½Ã½ºÅÛ

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) µ¿ÀÛ ÀνÄÀ» À§ÇÑ ±³»ç-Çлý ±¸Á¶ ±â¹Ý CNN
¿µ¹®Á¦¸ñ(English Title) Teacher-Student Architecture Based CNN for Action Recognition
ÀúÀÚ(Author) Yulan Zhao   ÀÌÈ¿Á¾   Yulan Zhao   Hyo Jong Lee  
¿ø¹®¼ö·Ïó(Citation) VOL 11 NO. 03 PP. 0099 ~ 0104 (2022. 03)
Çѱ۳»¿ë
(Korean Abstract)
´ëºÎºÐ ÷´Ü µ¿ÀÛ ÀÎ½Ä ÄÁº¼·ç¼Ç ³×Æ®¿öÅ©´Â RGB ½ºÆ®¸²°ú ±¤ÇÐ È帧 ½ºÆ®¸², ¾ç ½ºÆ®¸² ¾ÆÅ°ÅØó¸¦ ±â¹ÝÀ¸·Î ÇÏ°í ÀÖ´Ù. RGB ÇÁ·¹ÀÓ ½ºÆ®¸²Àº ¸ð¾ç Ư¼ºÀ» ³ªÅ¸³»°í ±¤ÇÐ È帧 ½ºÆ®¸²Àº µ¿ÀÛ Æ¯¼ºÀ» Çؼ®ÇÑ´Ù. ±×·¯³ª ±¤ÇÐ È帧Àº °è»ê ºñ¿ëÀÌ ¸Å¿ì ³ô±â ¶§¹®¿¡ µ¿ÀÛ ÀÎ½Ä ½Ã°£¿¡ Áö¿¬À» ÃÊ·¡ÇÑ´Ù. ÀÌ¿¡ ¾ç ½ºÆ®¸² ³×Æ®¿öÅ©¿Í ±³»ç-Çлý ¾ÆÅ°ÅØó¿¡¼­ ¿µ°¨À» ¹Þ¾Æ Çൿ ÀνÄÀ» À§ÇÑ »õ·Î¿î ³×Æ®¿öÅ© µðÀÚÀÎÀ» °³¹ßÇÏ¿´´Ù. Á¦¾È ½Å°æ¸ÁÀº µÎ °³ÀÇ ÇÏÀ§ ³×Æ®¿öÅ©·Î ±¸¼ºµÇ¾îÀÖ´Ù. Áï, ±³»ç ¿ªÇÒÀ» ÇÏ´Â ±¤ÇÐ È帧 ÇÏÀ§ ³×Æ®¿öÅ©¿Í Çлý ¿ªÇÒÀ» ÇÏ´Â RGB ÇÁ·¹ÀÓ ÇÏÀ§ ³×Æ®¿öÅ©¸¦ ¿¬°áÇÏ¿´´Ù. ÈÆ·Ã ´Ü°è¿¡¼­ ±¤ÇÐ È帧ÀÇ Æ¯Â¡À» ÃßÃâÇÏ°í ±³»ç ¼­ºê ³×Æ®¿öÅ©¸¦ ÈƷýÃŲ ´ÙÀ½ ±× Ư¡À» Çлý ¼­ºê ³×Æ®¿öÅ©¸¦ ÈƷýÃÅ°±â À§ÇÑ ±âÁؼ±À¸·Î ÁöÁ¤ÇÏ¿© Çлý ¼­ºê ³×Æ®¿öÅ©¿¡ Àü¼ÛÇÑ´Ù. Å×½ºÆ® ´Ü°è¿¡¼­´Â ±¤ÇÐ È帧À» °è»êÇÏÁö ¾Ê°í ´ë±â ½Ã°£ÀÌ ÁÙ¾îµéµµ·Ï Çлý ³×Æ®¿öÅ©¸¸ »ç¿ëÇÑ´Ù. Á¦¾È ³×Æ®¿öÅ©´Â ½ÇÇèÀ» ÅëÇÏ¿© Á¤È®µµ ¸é¿¡¼­ ÀÏ¹Ý ÀÌÁß ½ºÆ®¸² ¾ÆÅ°ÅØó¿¡ ºñÇØ ³ôÀº Á¤È®µµ¸¦ º¸¿©ÁÖ´Â °ÍÀ» È®ÀÎÇÏ¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Convolutional neural network (CNN) generally uses two-stream architecture RGB and optical flow stream for its action recognition function. RGB frames stream display appearance and optical flow stream interprets its action. However, the standard method of using optical flow is costly in its computational time and latency associated with increased action recognition. The purpose of the study was to evaluate a novel way to create a two sub-networks in neural networks. The optical flow sub-network was assigned as a teacher and the RGB frames as a student. In the training stage, the optical flow sub-network extracts features through the teacher sub-network and transmits the information to student sub-network for baseline training. In the test stage, only student sub-network was operational with decreased in latency without computing optical flow. Experimental results shows that our network fed only by RGB stream gets a competitive accuracy of 54.5% on HMDB51, which is 1.5 times better than that on R3D-18.
Å°¿öµå(Keyword) ¾ç ½ºÆ®¸²   ±³»ç-Çлý ¾ÆÅ°ÅØó   CNN   ±¤ÇÐ È帧   µ¿ÀÛ ÀνĠ  Two-Stream Network   Teacher-Student Architecture   CNN   Optical Flow   Action Recognition  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå