티스토리 뷰

[업데이트 2018.09.08 17:19]


1. 베이스 논문

[논문 요약17] Perceptual Generative Adversarial Networks for Small Object Detection

(http://arclab.tistory.com/172)


Jin et al. [19] proposed to train the CNN with hingle loss, which provides better test accuracy and faster stable convergence.


2. 레퍼런스 논문

[19] J. Jin, K. Fu, and C. Zhang. Traffic sign recognition with hinge loss trained convolutional neural networks. IEEE Transactions on Intel- ligent Transportation Systems, 15(5):1991–2000, 2014. 2

(https://ieeexplore.ieee.org/document/6766231/)


3. 주요 내용 요약

3.1 레퍼런스 논문 주요 내용

레퍼런스 논문에서는 hinge loss stochastic gradient descent (HLSGD)를 사용하여 convolutional neural networks (CNNs)를 학습하는 방법을 제안하고 있습니다. Hinge loss를 사용함으로써 테스트 성능의 정확도를 높이고, 학습시 빠르고 안정적으로 global minimum을 찾아 수렴할 수 있다고 합니다.




3.2 베이스 논문에서 인용한 내용

Related work관련하여 기존에 Traffic sign detection 문제의 성능 향상을 위해 hinge loss를 사용하는 CNN에 대해 언급하고 있습니다.



댓글
공지사항
최근에 올라온 글
최근에 달린 댓글
Total
Today
Yesterday
링크
«   2024/05   »
1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30 31
글 보관함