Á¤º¸Ã³¸®ÇÐȸ ³í¹®Áö ¼ÒÇÁÆ®¿þ¾î ¹× µ¥ÀÌÅÍ °øÇÐ
ÇѱÛÁ¦¸ñ(Korean Title) |
À̹ÌÁö À̾îºÙÀ̱⸦ ÀÌ¿ëÇÑ Àΰ£-°´Ã¼ »óÈ£ÀÛ¿ë ŽÁö µ¥ÀÌÅÍ Áõ° |
¿µ¹®Á¦¸ñ(English Title) |
Human-Object Interaction Detection Data Augmentation Using Image Concatenation |
ÀúÀÚ(Author) |
ÀÌ»ó¹é
À̱Ôö
Sang-Baek Lee
Kyu-Chul Lee
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 12 NO. 02 PP. 0091 ~ 0098 (2023. 02) |
Çѱ۳»¿ë (Korean Abstract) |
Àΰ£-°´Ã¼ »óÈ£ÀÛ¿ë ŽÁö´Â °´Ã¼ ŽÁö¿Í »óÈ£ÀÛ¿ë ÀνÄÀ» ÇÔ²² Ç®¾î¾ßÇÏ´Â ºÐ¾ß·Î ŽÁö ¸ðµ¨ÀÇ ÇнÀÀ» À§Çؼ ¸¹Àº µ¥ÀÌÅ͸¦ ÇÊ¿ä·Î ÇÑ´Ù. ÇöÀç °ø°³µÈ µ¥ÀÌÅͼÂÀº ±Ô¸ð°¡ ºÎÁ·ÇÏ¿© µ¥ÀÌÅÍ Áõ° ±â¹ý¿¡ ´ëÇÑ ¿ä±¸°¡ Ä¿Áö°í ÀÖÀ¸³ª, ´ëºÎºÐÀÇ ¿¬±¸¿¡¼ ±âÁ¸ÀÇ °´Ã¼ ŽÁö, À̹ÌÁö ºÐÇÒ ºÐ¾ß¿¡¼ È°¿ëÇÏ´Â Áõ° ±â¹ýÀ» È°¿ëÇÏ°í ÀÖ´Â ½ÇÁ¤ÀÌ´Ù. ÀÌ¿¡ º» ¿¬±¸¿¡¼´Â Àΰ£-°´Ã¼ »óÈ£ÀÛ¿ë ŽÁö ºÐ¾ß¿¡¼ È°¿ëÇÏ´Â µ¥ÀÌÅͼÂÀÇ Æ¯¼ºÀ» ÆľÇÇÏ°í, À̸¦ ÅëÇØ Àΰ£-°´Ã¼ »óÈ£ÀÛ¿ë ŽÁö ¸ðµ¨ ¼º´É Çâ»ó¿¡ È¿°úÀûÀÎ µ¥ÀÌÅÍ Áõ° ±â¹ýÀ» Á¦¾ÈÇÑ´Ù. º» ¿¬±¸¿¡¼ Á¦¾ÈÇÑ Áõ° ±â¹ý¿¡ ´ëÇÑ °ËÁõÀ» À§ÇÏ¿© ½ÇÇè ȯ°æÀ» ±¸ÃàÇÏ°í, ±âÁ¸ÀÇ ÇнÀ ¸ðµ¨¿¡ Àû¿ëÇÏ¿© Áõ° ±â¹ýÀ» Àû¿ëÇÒ °æ¿ì¿¡ ŽÁö ¸ðµ¨ÀÇ ¼º´É Çâ»óÀÌ °¡´ÉÇÔÀ» È®ÀÎÇÏ¿´´Ù. |
¿µ¹®³»¿ë (English Abstract) |
Human-object interaction(HOI) detection requires both object detection and interaction recognition, and requires a large amount of data to learn a detection model. Current opened dataset is insufficient in scale for training model enough. In this paper, we propose an easy and effective data augmentation method called Simple Quattro Augmentation(SQA) and Random Quattro Augmentation(RQA) for human-object interaction detection. We show that our proposed method can be easily integrated into State-of-the-Art HOI detection models with HICO-DET dataset. |
Å°¿öµå(Keyword) |
Àΰ£-°´Ã¼ »óÈ£ÀÛ¿ë ŽÁö
µ¥ÀÌÅÍ Áõ°
À̹ÌÁö ó¸®
°´Ã¼ ŽÁö
Human-object Interaction Detection
Data augmentation
Image Processing
Object Detection
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|