IPIU 2016 (Á¦28ȸ ¿µ»óó¸® ¹× ÀÌÇØ¿¡ °üÇÑ ¿öÅ©¼¥)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
Finding Face-Lookalikes Using CNN |
¿µ¹®Á¦¸ñ(English Title) |
Finding Face-Lookalikes Using CNN |
ÀúÀÚ(Author) |
Junho Cho
Wangue Jee
YoungBae Jeon
Sangdoo Yun
Jin Young Choi
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 28 NO. 01 PP. P1 ~ 0037 (2016. 02) |
Çѱ۳»¿ë (Korean Abstract) |
|
¿µ¹®³»¿ë (English Abstract) |
Face-lookalikes intrigue people and are common subjects in our daily conversation. Finding face-lookalikes but has not been a popular topic in computer vision compared to face verification/identification. Resemblance in faces are subjective to every individual, and the absence of absolute measure of it makes finding face-lookalikes difficult. To the best of our knowledge, this is the first work of finding face-lookalikes. We make two contributions: first, we propose a procedure to make face retrieval dataset for finding face-lookalikes by crawling faces on web. Second, we propose a user-friendly framework of retrieving top-k face-lookalikes to the user¡¯s device. We evaluate our experiment results on our finely webcrawled dataset, suggesting more lookalike faces than other public dataset. |
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|