°ÇÃ൵½Ã°ø°£¿¬±¸¼Ò

Architecture & Urban Research Institute

pdf¿ø¹®º¸±â ¿¡·¯ ÇØ°á¹æ¹ý ¹Ù·Î°¡±â



¹®ÇåȨ > ¿¬±¸³í¹® > »ó¼¼

[¿ø¹®º¸±â½Ã ¼ÒºñµÇ´Â Æ÷ÀÎÆ® : 100 Æ÷ÀÎÆ®] ¹Ì¸®º¸±â Àοë

Çѱ¹°ø°£µðÀÚÀÎÇÐȸ|³í¹®Áý 2023³â 4¿ù

³í¹®¸í µö·¯´× ±â¹Ý Áö¿ª »ó¾÷°¡·ÎÀÇ ÆÄ»çµå µðÀÚÀÎ ¾ÆÀ̵§Æ¼Æ¼ ±¸º° - ¼­¿ï ¼º¼öµ¿ µî »ó¾÷°¡·Î ·Îµåºä À̹ÌÁöµ¥ÀÌÅÍ ÇнÀ¸ðµ¨À» ±â¹ÝÀ¸·Î - / Distinguishing Facade Design Identity from Local Commercial Street Based on Deep Learning - Focusing on street view image training model in some areas including Seongsu-dong, Seoul -
ÀúÀÚ¸í Á¶ÇÏ¿µ ; ÀÌÁø±¹
¹ßÇà»ç Çѱ¹°ø°£µðÀÚÀÎÇÐȸ
¼ö·Ï»çÇ× Çѱ¹°ø°£µðÀÚÀÎÇÐȸ ³í¹®Áý, Vol.18 No.03 (2023-04)
ÆäÀÌÁö ½ÃÀÛÆäÀÌÁö(305) ÃÑÆäÀÌÁö(10)
ISSN 1976-4405
ÁÖÁ¦ºÐ·ù °èȹ¹×¼³°è / µµ½Ã
ÁÖÁ¦¾î µö·¯´×; Áö¿ª »ó¾÷°¡·Î; ÆÄ»çµå; ¾ÆÀ̵§Æ¼Æ¼; ·Îµåºä ; Deep Learning; Local Commercial street; Facade; Identity; Street view
¿ä¾à1 (¿¬±¸¹è°æ ¹× ¸ñÀû) º» ³í¹®Àº ¼­¿ïÀÇ ¿©·¯ Áö¿ª »ó¾÷°¡·Î¿¡ À§Ä¡ÇÑ °ÇÃ๰ ÆÄ»çµå À̹ÌÁö¸¦ È°¿ëÇÏ¿© µö·¯´× ÇнÀ¸ðµ¨À» ±¸ÃàÇÔÀ¸·Î½á Áö¿ªº° ÆÄ»çµå ¾ÆÀ̵§Æ¼Æ¼ ±¸º° Á¢±Ù ¹æ¾È¿¡ °üÇØ ±â¼úÇÏ°íÀÚ ÇÑ´Ù. Áö¿ª¿¡ À־ ¾ÆÀ̵§Æ¼Æ¼´Â °³ÀÎÀÇ ÀÎÁö¿Í °ü·ÃµÈ °ÍÀ¸·Î, °³ÀÎÀÌ Áö¿ª¿¡ ´ëÇØ ÀνÄÇÏ°í ÀÖ´Â ÃÑüÀûÀÎ À̹ÌÁöÀÇ ÁýÇÕÀ̶ó°í ÇÒ ¼ö ÀÖÀ¸¸ç, ±¹³» ¼­¿ï ¿©·¯ Áö¿ª¿¡¼­´Â ±Þ¼ÓÇÑ µµ½ÃÈ­ °úÁ¤ Áß Áö¿ªÀÇ ¾ÆÀ̵§Æ¼Æ¼°¡ º¯È­ ¹× ¼Ò½Ç, È®¸³µÇ±âµµ Çϸç ÀÌ¿Í °ü·ÃµÈ ¿¬±¸°¡ À̾îÁ® ¿À°í ÀÖ´Ù. º» ¿¬±¸´Â ÀÌ·¯ÇÑ Áö¿ª ¾ÆÀ̵§Æ¼Æ¼¸¦ ±¸¼ºÇÏ´Â ´Ù¾çÇÑ ±¸¼º ¿ä¼Ò Áß ½Ã°¢Àû ¿ä¼Ò¿¡ ÃÊÁ¡À» ¸ÂÃß¾î, µö·¯´× ±â¼úÀ» È°¿ëÇÏ¿© Áö¿ªÀÇ ½Ã°¢Àû µ¥ÀÌÅ͸¸À¸·Î °¢ Áö¿ªÀÇ ¾ÆÀ̵§Æ¼Æ¼°¡ Ãß·ÐµÉ ¼ö ÀÖ´ÂÁö È®ÀÎÇÑ´Ù. ÃÖ±Ù µö·¯´× ±â¼úÀº ÀΰøÁö´É¿¡ ´ëÇÑ ÀνÄÀÌ ³ô¾ÆÁü¿¡ µû¶ó ¿µ»ó, À̹ÌÁö, ÅؽºÆ® µî ´Ù¾çÇÑ ÄÁÅÙÃ÷¿¡ È®ÀåµÇ¾î »ç¿ëµÇ°í ÀÖÀ¸¸ç ƯÈ÷ °ÇÃà ¹× °ÇÃà°ø°£ µðÀÚÀο¡ ÀÖ¾î ½Ã°¢Áö´É¸ðµ¨À» È°¿ëÇÑ ´Ù¾çÇÑ ±â°èÇнÀ ±â¹ÝÀÇ Ãß·Ð ¹× ÀÚµ¿È­ µî ÀÀ¿ë ¿¬±¸°¡ À̾îÁ® ¿À°í ÀÖ´Ù. (¿¬±¸¹æ¹ý) º» ¿¬±¸´Â ¼­¿ï½Ã ³» ¿©·¯ »ó¾÷°¡·Î¸¦ ´ë»óÀ¸·Î µ¥ÀÌÅÍ ¼öÁýÀ» ÁøÇàÇÏ°í ¼öÁýÇÑ µ¥ÀÌÅ͸¦ °¡°øÇÏ¿© µö·¯´× ¸ðµ¨À» ±¸ÃàÇÑ´Ù. ÀÌ¿¡ ´ëÇÑ ±¸Ã¼Àû °úÁ¤À» ¿ä¾àÇÏÀÚ¸é 1) Æ÷ÅлçÀÌÆ®ÀÇ ·Îµåºä ±â´ÉÀ» È°¿ëÇÑ Áö¿ªº° °ÇÃ๰ ÆÄ»çµå À̹ÌÁö ȹµæ, 2) À̹ÌÁö µ¥ÀÌÅÍ Àü󸮸¦ ÅëÇÑ ÆÄ»çµå µ¥ÀÌÅͼ¼Æ® ±¸Ãà, 3) »çÀü ÈÆ·ÃµÈ CNN ±â¹ÝÀÇ µö·¯´× ¸ðµ¨ È°¿ë ¹× ÇнÀ, ±×¸®°í 4) Å×½ºÆ® À̹ÌÁö µ¥ÀÌÅ͸¦ ÅëÇÑ Å×½ºÆ® °á°ú È®ÀÎ ¹× ³íÀÇ ¼øÀ¸·Î ÁøÇàÇÑ´Ù. (°á°ú) ¿¬±¸ °á°ú, ±¸ÃàÇÑ Áö¿ªº° ÆÄ»çµå À̹ÌÁö µ¥ÀÌÅͼ¼Æ® ¹× µö·¯´× ÇнÀ¸ðµ¨À» ÅëÇØ °¢ »ó¾÷°¡·ÎÀÇ ÆÄ»çµå À̹ÌÁö¶ó´Â ½Ã°¢Á¤º¸¸¸À¸·Îµµ °¢ Áö¿ªº°·Î À̹ÌÁö°¡ ºÐ·ùµÇ´Â °ÍÀ» È®ÀÎÇÏ¿´´Ù. ƯÈ÷, ÇØ´ç ¸ðµ¨ÀÇ ÇнÀ ¹× Å×½ºÆ® °á°ú ƯÁ¤ Áö¿ªÀÇ ÆÄ»çµå À̹ÌÁö°¡ ³ôÀº Á¤È®µµ·Î ºÐ·ùµÇ´Â ¸ð½ÀÀ» È®ÀÎÇÒ ¼ö ÀÖ¾ú´Ù. ÀÌ´Â »çÀüµµ¸ÞÀÎ Áö½ÄÀ» ÅëÇØ ÇØ´ç Áö¿ªÀÌ Å¸ Áö¿ªµé¿¡ ºñÇØ ÇÁ·£Â÷ÀÌÁî ¸ÅÀåÀÇ ºñÀ²ÀÌ ¶Ñ·ÇÇÏ°Ô ³·Àº µî ÇØ´ç »ó¾÷°¡·Î¿¡ ´ëÇÑ Á¤Ã¥ µîÀÇ ÀÌÀ¯°¡ ±Ù°ÅÀÎ °ÍÀ¸·Î ÃßÁ¤µÉ ¼ö ÀÖÀ¸¸ç À̸¦ È°¿ëÇÏ¿© ½É¹ÌÀûÀÎ ¿ä¼Ò¸¸À¸·Î Áö¿ªº° »ó¾÷°¡·ÎÀÇ ÆÄ»çµå À̹ÌÁö ÇнÀ¸ðµ¨À» ÅëÇØ »ó¾÷°¡·Î ÆÄ»çµåº° ¾ÆÀ̵§Æ¼Æ¼°¡ Ãß·Ð °¡´ÉÇÔÀ» È®ÀÎÇÒ ¼ö ÀÖ¾ú´Ù. (°á·Ð) ±âÁ¸ÀÇ Áö¿ª ¹× »ó¾÷°¡·ÎÀÇ ¾ÆÀ̵§Æ¼Æ¼ °ü·Ã ¿¬±¸´Â ÁÖ·Î Áö¿ª¹Îµé, ¹æ¹®°´µéÀ» ´ë»óÀ¸·Î ÀÎÅÍºä ¹× ¼³¹® Á¶»ç µî Á¤¼ºÀûÀÎ ¹æ¹ý·ÐÀ» È°¿ëÇÏ¿© Áö¿ª¼ºÀ» Á¤ÀÇ ³»·Á¿ÔÀ¸³ª, º» ¿¬±¸ÀÇ Á¢±Ù ¹æ¾ÈÀ» ÅëÇØ È®ÀÎÇÑ °á°ú¸¦ È°¿ëÇÏ¿© º¸´Ù ½Ã°£Àû, ºñ¿ëÀûÀ¸·Î È¿À²ÀûÀÎ Á¤·®Àû Á¢±Ù ¹æ¾ÈÀ» Á¦½ÃÇÏ¿´´Ù´Â µ¥ ÀÇÀÇ°¡ ÀÖÀ¸¸ç, º» ¿¬±¸¸¦ ÅëÇØ ÃßÈÄ ¿©·¯ Áö¿ªÀÇ µ¥ÀÌÅÍ ¼öÁý ¹× °¡°øÀ» ±â¹ÝÀ¸·Î Áö¿ªº° °æ°ü°èȹ ½Ã ±âÃÊ ¿¬±¸·Î½á Á¦°øµÉ °ÍÀ¸·Î ±â´ëÇÑ´Ù.
¿ä¾à2 (Background and Purpose) This paper describes an approach to distinguishing the identity of facades by region. The study constructed a deep learning model using facade images of buildings located on various local commercial streets in Seoul. In each region, identity is related to individual perceptions and can be described as a collection of images that individuals perceive regarding that region. In many parts of Seoul, local identities have changed, been lost, or have been established during the rapid urbanization process, and related research has been conducted. This study focused on visual elements among the various components that make up these local identities, and utilized deep learning technology to confirm whether each region's identity can be inferred from local visual data alone. With the increased awareness of artificial intelligence, deep learning technology has been expanded and used in various contexts, such as creating images and text, and research on applications such as inference and automation using visual intelligence models has continued in the architectural spatial design field. (Method) This study collected data from various commercial streets in Seoul and built a deep learning model by processing the collected data. The specific processes can be summarized in the following order: 1) acquisition of regional building facade images using the street view function of portal sites, 2) construction of facade datasets through image data preprocessing, 3) utilization and learning of pre-trained CNN-based deep learning models, and 4) a review and discussion of the results from the test image data. (Results) Through the built regional facade image dataset and deep learning model, it was confirmed that images were classified by region only with visual information called facade images of each commercial street. In particular, as a result of learning and testing the model, it was confirmed that facade images in a specific region were classified with high accuracy. This can be presumed to be due to reasons such as policies for commercial streets in the region, such as a clearly lower proportion of franchise stores than in other regions. Thus, it was confirmed that identity can be inferred through the facade image learning model of commercial streets by region using only aesthetic factors. (Conclusions) Existing regional and commercial identity-related studies have defined locality using qualitative methodologies, such as interviews and surveys of local residents and visitors. However, this study is meaningful in that it presented a more time- and cost-effective quantitative approach.
¼ÒÀåó Çѱ¹°ø°£µðÀÚÀÎÇÐȸ
¾ð¾î Çѱ¹¾î
DOI http://10.35216/kisd.2023.18.3.305
¡á Ãßõ¹®Çå (ÀÌ ¹®Çå°ú °°ÀÌ º» ¹®Çå)
[½Ã·Ð] 4Â÷ »ê¾÷Çõ¸í ½Ã´ë, °ÇÃàÀÇ ´ëÀÀ
À̸í½Ä(Lee, Myung-Sik) - °ÇÃà(´ëÇÑ°ÇÃàÇÐȸÁö) : Vol.61 No.05 (201705)
ÇÁ·Î±×·¥°ú ´ÙÀ̾î±×·¥ÀÇ Àû¿ëÀ» ÅëÇÑ °ÇÃà ÇüÅ »ý¼º ¿¬±¸
À±Çý°æ(Yoon Hae-Kyung) ; ±èÁÖ¿µ(Kim Ju-Young) ; È«¿øÈ­(Hong Won-Hwa) - ´ëÇÑ°ÇÃàÇÐȸ³í¹®Áý °èȹ°è : v.26 n.02 (201002)
·½ ÄðÇϽº¿Í Ä«Áî¿ä ¼¼Áö¸¶ °ÇÃà¿¡¼­ ³ªÅ¸³ª´Â ÇÁ·Î±×·¥ÀÇ Á¶Á÷°ú °ø°£±¸¼º ¹æ¹ýÀÇ ºñ±³ºÐ¼®
±Ç°æ¹Î ; ±èÁ¾Áø - Çѱ¹½Ç³»µðÀÚÀÎÇÐȸ ³í¹®Áý : v.16 n.6(Åë±Ç 65È£) (200712)
¾Æ¸ð·¹ÆÛ½ÃÇÈ »ç¿Á ¼³°è»ç·Ê
ÃÖµµ¼® - ¼³ºñÀú³Î(´ëÇѼ³ºñ°øÇÐȸÁö) : Vol.47 No.12 (201812)
°ÇÃ༳°è´Ü°è Áö½Ä±â¹Ý ¸¶°¨ÀÚÀç ¼±Á¤ ÇÁ·ÎÅäŸÀÔ °³¹ß
À±¼öÈ£(Yun Su-Ho) ; ¹ÚÇüÁø(Park Hyung-Jin) ; ±¸±³Áø(Koo Kyo-Jin) - ´ëÇÑ°ÇÃàÇÐȸ³í¹®Áý ±¸Á¶°è : v.27 n.7 (201107)
½Ã·Ð: Äڷγª ÀÌÈÄ, °ÇÃà°ú µµ½Ã°ø°£ÀÇ º¯È­¸¦ »ý°¢ÇÏ´Ù
¹Ú¼ÒÇö - °ÇÃà°ú µµ½Ã°ø°£ : Vol.38 (202006)
¿ÀºêÁ§Æ®-ÆĶó¹ÌÅͱâ¹Ý °ÇÃึ°¨°ø»çºñ °³»ê°ßÀû ¸ðµ¨
±¸±³Áø(Koo, Kyo-Jin) ; ¹Ú¼ºÈ£(Park, Sung-Ho) ; ¹Ú¼ºÃ¶(Park, Sung-Chul) ; ¼ÛÁ¾°ü(Song, Jong-Kwan) - Çѱ¹°Ç¼³°ü¸®ÇÐȸ ³í¹®Áý : Vol.9 No.6 (200812)
[ƯÁý] ±¹³» ģȯ°æ °ÇÃ๰ »ç·Ê ºÐ¼®
Á¤Áö³ª ; ±è¿ë¼® ; À̽¹Π- ±×¸°ºôµù(Çѱ¹±×¸°ºôµùÇùÀÇȸÁö) : v.9 n.1 (200803)
ÆĶó¸ÞÆ®¸¯ µðÀÚÀÎ ±â¹ý°ú À¯Àü¾Ë°í¸®ÁòÀ» Àû¿ëÇÑ ¼º´É±â¹Ý °øµ¿ÁÖÅà °èȹ ¹æ¹ý ¿¬±¸
Á¤¿äÇÑ(Jeong, Yo-Han) ; ±Ç¿µ»ó(Kwon, Young-Sang) - ´ëÇÑ°ÇÃàÇÐȸ³í¹®Áý : Vol.37 No.12 (202112)
BIM ±â¹Ý ºñ¼±Á¤ ÀÛ¾÷Ç׸ñ ¹°·®»êÃâ ¹æ¹ý¿¡ °üÇÑ ¿¬±¸
¹Ú»óÇå;À±¼±Àç;±¸±³Áø - Çѱ¹°ÇÃà½Ã°øÇÐȸ Çмú.±â¼ú³í¹®¹ßǥȸ ³í¹®Áý : v.19 n.2(Åë±Ç Á¦37È£) (201911)