´Ý±â
Loading..

ÀüÀÚÁ¤º¸¿¬±¸Á¤º¸¼¾ÅÍ ICT À¶ÇÕ Àü¹®¿¬±¸Á¤º¸ÀÇ Áý´ë¼º

Çмú´ëȸ ÇÁ·Î½Ãµù

Ȩ Ȩ > ¿¬±¸¹®Çå > Çмú´ëȸ ÇÁ·Î½Ãµù > Çѱ¹Á¤º¸°úÇÐȸ Çмú´ëȸ > KCC 2020

KCC 2020

Current Result Document : 0 / 0

ÇѱÛÁ¦¸ñ(Korean Title) ÁúÀǸ¦ Ȱ¿ëÇÑ Á¦·Î¼¦ °´Ã¼ ŽÁö
¿µ¹®Á¦¸ñ(English Title) Query-based zero-shot detection
ÀúÀÚ(Author) ¼ÛÈ£ÁØ   À̰­ÈÆ   À庴Ź   HoJoon Song   Ganghun Lee   Byoung-Tak Zhang  
¿ø¹®¼ö·Ïó(Citation) VOL 47 NO. 01 PP. 0886 ~ 0888 (2020. 07)
Çѱ۳»¿ë
(Korean Abstract)
¿µ¹®³»¿ë
(English Abstract)
The major development in object classification and detection models has already reached a point to where we can use deep learning models in real-life applications. However, there still exists the infamous drawback: large-scale annotated data are required for training. As such, zero-shot learning, predictions of data that has not been trained, is gathering attention in the machine learning community. In this paper, we propose a zero-shot detection method using a query-based model to detect untrained objects. The proposed method makes use of an extra image(the query) of the target object for the detection task. We verified the proposed method through experiments using handwritten characters. The results showed that the proposed method can successfully detect the bounding boxes of untrained objects.
Ű¿öµå(Keyword)
¿ø¹® PDF ´Ù¿î·Îµå