search

mingxoxo / open-cv

2 stars 1 forks source link

논문 공부 #11

Open mingxoxo opened 5 years ago

mingxoxo commented 5 years ago

Siamese Neural Networks for One-shot Image Recognition(샴 네트워크) : https://jayhey.github.io/deep%20learning/2018/02/06/saimese_network/ https://www.edwith.org/deeplearningai4/lecture/34912/ loss function : https://blog.naver.com/infoefficien/221254432488 Siamese기반 추적 기법 -> 빠른 속도 , 낮은 성능 낮은 성능 -> 첫 프레임의 물체 외형으로 현재 프레임에서의 물체를 찾는 것으로 외형변화를 무시하기 때문에 외형변화가 심할 때 물체 추적을 실패하는 경우가 많음

SiamFC : http://openresearch.ai/t/siamesefc-fully-convolutional-siamese-networks-for-object-tracking/200 AlexNet https://dalpo0814.tistory.com/23 CNN : http://taewan.kim/post/cnn/#fnref:1 Batch normalization: https://sacko.tistory.com/44?category=632408

mingxoxo commented 5 years ago

image https://sacko.tistory.com/37

mingxoxo commented 5 years ago

Pooling : Convolution을 거쳐서 나온 activation maps이 있을 때, 이를 이루는 convolution layer을 resizing하여 새로운 layer를 얻는 것 Max pooling : 최댓값을 갖는 pooling image https://blog.naver.com/varkiry05/221065577703 https://medium.com/@hobinjeong/cnn%EC%97%90%EC%84%9C-pooling%EC%9D%B4%EB%9E%80-c4e01aa83c83

mingxoxo commented 5 years ago

image https://en.wikipedia.org/wiki/Loss_functions_for_classification#Logistic_loss

mingxoxo commented 5 years ago

1

2

3