Ȩ > Ç¥ÁØÈ °³¿ä > TTAÀÇ Ç¥ÁØÇöȲ
Ç¥ÁعøÈ£ | TTAK.KO-11.0230 | ±¸Ç¥ÁعøÈ£ | |
---|---|---|---|
Á¦°³Á¤ÀÏ | 2017-12-13 | ÃÑÆäÀÌÁö | 15 |
ÇѱÛÇ¥Áظí | ÀÓº£µðµå ½Ã½ºÅÛÀ» À§ÇÑ ½Ã°¢ Á¤º¸ ó¸® ¼³°è ±¸Á¶ | ||
¿µ¹®Ç¥Áظí | Visual Information Processing Design Architecture for Embedded Systems | ||
Çѱ۳»¿ë¿ä¾à | º» Ç¥ÁØÀº ÀÓº£µðµå ½Ã½ºÅÛ¿¡¼ ½Ã°¢ Á¤º¸ ó¸®¿Í ºñÀü ÀνÄÀ» À§ÇÑ OpenVX ±â¹ÝÀÇ ½Ã°¢ Á¤º¸ ó¸® ¼³°è ±¸Á¶¿Í À̹ÌÁö ÀÎÅÍÆäÀ̽º¸¦ Á¦½ÃÇÑ´Ù. ÀÓº£µðµå ½Ã½ºÅÛ¿¡¼´Â Çϵå¿þ¾î¿Í ¼ÒÇÁÆ®¿þ¾îÀÇ Á¦¾à »çÇ×À¸·Î ÀÎÇØ ±âÁ¸ÀÇ µ¥½ºÅ©ÅéÀ̳ª ¼¹ö¸¦ ÀÌ¿ëÇÑ ½Ã°¢ Á¤º¸ ó¸® ±â¼úµéÀ» ±×´ë·Î »ç¿ëÇؼ´Â ÃÖÀûÀÇ ¼º´ÉÀ» È®º¸Çϱ⠾î·Æ´Ù. ¶ÇÇÑ ¸Ó½Å·¯´×À» À§ÇÑ ÀÔ·Â À̹ÌÁö´Â À̹ÌÁöÀÇ Àνİú ºÐ·ù °úÁ¤¿¡¼ Á¤È®¼º°ú È¿À²¼ºÀ» ³ôÀ̱â À§ÇØ ±âÃÊÀûÀÎ Àüó¸® °úÁ¤ÀÌ ¿ä±¸µÇ¸ç, ÀÌ´Â ¸¹Àº ÄÄÇ»Æà ÀÚ¿øÀ» ÇÊ¿ä·Î ÇÑ´Ù. À̸¦ À§ÇØ º» Ç¥ÁØ¿¡¼´Â ±¸Á¶ÀûÀ¸·Î ½Ã½ºÅÛÀÇ ·ÎÄà ¸Þ¸ð¸®¿Í ij½¬ ¸Þ¸ð¸®¸¦ ÃÖ´ëÇÑ È°¿ëÇÏ°í, ºÒÇÊ¿äÇÑ º¹»ç¸¦ Á¦°ÅÇÏ¿© ÀûÀº ÄÄÇ»Æà ÀÚ¿øÀ» ÅëÇØ ÃÖ´ëÇÑÀÇ ºñÀü 󸮸¦ Áö¿øÇÒ ¼ö ÀÖ´Â OpenVX ÇÁ·¹ÀÓ¿öÅ© Ç¥ÁØ ¸í¼¼(specification)¿¡ µû¶ó ½Ã°¢ Á¤º¸ 󸮸¦ À§ÇÑ ±âº» ±¸Á¶¿Í ÀÎÅÍÆäÀ̽º¸¦ ±â¼úÇÑ´Ù. | ||
¿µ¹®³»¿ë¿ä¾à | The standard provides an OpenVX-based visual information processing design architecture and image interface for visual information processing and vision recognition in embedded system. In the systems, it is difficult to obtain optimum performance by using visual information processing technologies using existing desktop or server due to hardware and software limitations. Furthermore, the input image for machine learning requires a basic preprocessing process in order to improve accuracy and efficiency in image recognition and classification, which requires a lot of computing resources. To this purpose, we describe a design architecture and a interface for visual information processing in accordance with the OpenVX framework specification that can structurally maximize the system's local and cache memory, and eliminate unnecessary copying to support maximum vision processing through fewer computing resources. | ||
±¹Á¦Ç¥ÁØ | |||
°ü·ÃÆÄÀÏ |
![]() |
- ÀÌÀü
- È®À帮¼Ò½º ±â¼úÀÚ
- ´ÙÀ½
- ÄÁÆÛ·±½Ì ½Ã³ª¸®¿À