티스토리 뷰

[업데이트 2018.09.05 10:55]


1. 베이스 논문

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

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


We compare the result of Perceptual GAN with all the existing methods that achieved best performance on the Caltech testing set, including VJ [37], HOG [5], LDCF [27], Katamari [2], SpatialPooling+ [30], TA-CNN [36], Checkerboards [43], CompACT-Deep [44] and RPN+BF [41].


2. 레퍼런스 논문

[2] R. Benenson, M. Omran, J. Hosang, and B. Schiele. Ten years of pedestrian detection, what have we learned? In ECCV, pages 613– 627, 2014. 6

(https://arxiv.org/pdf/1411.4304.pdf)


3. 주요 내용 요약

3.1 레퍼런스 논문 주요 내용

- Small object detection(Pedestrian Detection)을 연구할 떄 작은 사이즈의 보행자 이미지가 있기 때문에 Caltech-USA dataset을 사용하는데, 본 논문에서는 10년간 나왔던 object detection 방법들중 일부를 조합하여 성능 향상을 시도하였고, 제안한 방법인 Katamari object detector를 통해 Caltech-USA dataset에사 가장 좋은 결과를 보여주었습니다.


- Katamari object detector : SquaresChnFtrs+DCT+SDt+2Ped 조합


3.2 베이스 논문에서 인용한 내용

Perceptual Generative Adversarial Networks for Small Object Detection 논문에서 제안한 Perceptual GAN이 Caltech-USA dataset에 대해 본 논문에서 제안한 방법인 Katamari object detector(SquaresChnFtrs+DCT+SDt+2Ped 조합)보다 성능이 좋음을 보여주고 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