• 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 : 584 / 585

ÇѱÛÁ¦¸ñ(Korean Title) ÇÕ¼º µ¥ÀÌÅ͸¦ ÅëÇÑ ºÎºÐ °¡·ÁÁü¿¡ °­ÀÎÇÑ ±º¿ë Â÷·® °ËÃâ
¿µ¹®Á¦¸ñ(English Title) Robust Military Vehicle Detection under Partial Occlusion with Synthetic Data
ÀúÀÚ(Author) Á¶¼±¿µ   Sunyoung Cho  
¿ø¹®¼ö·Ïó(Citation) VOL 27 NO. 11 PP. 0519 ~ 0530 (2021. 11)
Çѱ۳»¿ë
(Korean Abstract)
ÃÖ±Ù ½ÉÃþ ½Å°æ¸Á ±â¹Ý °´Ã¼ °ËÃâ ±â¼úÀÇ ¹ßÀü¿¡µµ ºÒ±¸ÇÏ°í ºÎºÐÀûÀ¸·Î °¡·ÁÁø °´Ã¼¸¦ °ËÃâÇÏ´Â °ÍÀº ¿©ÀüÈ÷ ¾î·Á¿î ¹®Á¦ÀÌ´Ù. °´Ã¼ÀÇ ¿Ü°ü ¹× ÇüÅ¿¡ ´ëÇÑ Á¦ÇÑÀûÀÎ Á¤º¸·Î ÀÎÇØ °¡·ÁÁüÀÌ ÀÖ´Â °´Ã¼¿¡ ´ëÇÑ Á¤È®ÇÑ ¹Ù¿îµù ¹Ú½º¸¦ ã°Å³ª Ŭ·¡½º¸¦ ±¸º°ÇÏ´Â °ÍÀÌ ¾î·Æ±â ¶§¹®ÀÌ´Ù. º» ³í¹®¿¡¼­´Â °¡·ÁÁüÀ» °®´Â µ¥ÀÌÅ͸¦ ÇÕ¼ºÇÏ¿© »ý¼ºÇÏ°í, À̸¦ ÀÌ¿ëÇÑ ¸ðµ¨ ÇнÀÀ» ÅëÇØ ºÎºÐ °¡·ÁÁüÀÌ ÀÖ´Â °´Ã¼ÀÇ °ËÃâ ¼º´ÉÀ» Çâ»ó½ÃÅ°´Â ¹æ¹ýÀ» Á¦¾ÈÇÑ´Ù. ´Ù¾çÇÑ °¡·ÁÁü »óȲÀ» °í·ÁÇϱâ À§ÇØ ´Ù¾çÇÑ °¡·ÁÁü ·¹º§ ¹× Á¾·ù¿¡ µû¶ó ÇÕ¼º µ¥ÀÌÅ͸¦ »ý¼ºÇÑ´Ù. Á¦¾ÈÇÏ´Â ¹æ¹ýÀÇ ¼º´ÉÀ» Æò°¡Çϱâ À§ÇØ ½ÇÁ¦ ±º¿ë Â÷·®¿¡ ´ëÇÑ µ¥ÀÌÅͼÂÀ» ¼öÁýÇÏ¿´°í, ÀÌ¿¡ ´ëÇÑ ÇÕ¼º µ¥ÀÌÅ͸¦ »ý¼ºÇÏ¿© ¸ðµ¨ ÇнÀ¿¡ È°¿ëÇÏ¿´´Ù. ´Ù¾çÇÑ ½ÇÇèÀ» ÅëÇØ ÇÕ¼º µ¥ÀÌÅ͸¦ ÀÌ¿ëÇÏ¿© ÇнÀÇÑ ¸ðµ¨ÀÌ ºÎºÐ °¡·ÁÁüÀ» °®´Â °´Ã¼ °ËÃâ ¼º´ÉÀ» Çâ»ó½ÃÅ´À» º¸¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Although advances in object detection are based on deep neural networks, detecting partially occluded objects remains a difficult task. Localizing or classifying objects under partial occlusion is difficult due to limited information about the appearances and shapes of the objects. This paper generates synthetically occluded data and presents a method to improve object detection under partial occlusion by synthetic data. We generated synthetic data with various levels and types of occlusion to consider various occlusion situations. To evaluate our method, we collect a military vehicle dataset and exploit the synthetically occluded data generated by our method for model learning. We show that our model trained with synthetic data improves object detection under partial occlusion through various experiments.
Å°¿öµå(Keyword) ºÎºÐ °¡¸²   °´Ã¼ °ËÃâ   °¡·ÁÁü ÇÕ¼º   ±º¿ë Â÷·® °ËÃâ   ½ÉÃþ ÇÕ¼º°ö ½Å°æ¸Á   partially occluded   object detection   synthetically occluded   military vehicle detection   deep convolutional neural networks  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå