ÇѱÛÁ¦¸ñ(Korean Title) |
A Study on specific region detection and recognition based on Transfer-Learning and Data Augmentation |
¿µ¹®Á¦¸ñ(English Title) |
A Study on specific region detection and recognition based on Transfer-Learning and Data Augmentation |
ÀúÀÚ(Author) |
Kwang-Seong Shin
Dong-Sik Jo
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 11 NO. 01 PP. 0384 ~ 0387 (2019. 06) |
Çѱ۳»¿ë (Korean Abstract) |
|
¿µ¹®³»¿ë (English Abstract) |
In this study, we study how to enhance data to perform deep learning using a small data set. Transfer learning refers to the use of Deep-Learning as a feature extractor and learning of other models with such extracted features. When you create a new model using the existing model, you have the advantage of speeding up the learning and increasing the prediction more. In this study, we apply a data augmentation technique that is used to solve the over sum problem that occurs after the transfer learning, and find a way to inflate the small data set itself and use it as learning data.
|
Å°¿öµå(Keyword) |
Transfer learning
Data augmentation
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|