Notice
Recent Posts
Recent Comments
Link
일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
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 |
Tags
- CNN구조정리
- 맥실리콘
- arm칩에안드로이드
- CS231n
- 백준
- professor strang
- CS231nAssignments
- 아이폰원스토어
- BOJ
- 맥북원스토어
- 선형대수학
- monoculardepthestimation
- ios원스토어
- RegionProposalNetworks
- Algorithm
- Gilbert Strang
- CS231nAssignment1
- Linear algebra
- CS231nSVM
- gpumemory
- adversarialattackonmonoculardepthestimation
- BAEKJOON
- CNNarchitecture
- pycharmerror
- 선대
- ㅐㅕ세ㅕㅅ
- CS231ntwolayerneuralnet
- MacOS
- MIT
- 백준알고리즘
Archives
- Today
- Total
개발로 하는 개발
[Encoder - Decoder] Architecture 본문
Image-level prediction
: classification -> output : probability vector class
( input 관계없이 output size가 정해져 있다)
Pixel-level prediction ( input : )
: segmentation -> output : probability volume
: denoising -> output : clean(denoised) image
( input 크기에 비례해 output 크기가 정해진다)
- Semantic Segmentation
-> pixel (instance X) 기준으로 class label을 정한다.
- 객체의 개수는 영향을 미치지 않고, 구분되지도 않는다.
- 단순히 같은 class끼리 labeling한다.
- sliding window
local region만 생각해서 그 window를 통째로 AlexNet에 통과시키는 방식
local한 범위 내에서의 prediction이 어려움
window가 겹치는 곳의 계산을 share하지 않아서 비효율적
- fully convolutional network
각 클래스(C개) 별 score를 pixel 단위로 전부 계산
original resolution에서 계산을 하게 되므로 expensive
- encoder - decoder network
downsampling & upsampling inside the network
-> upsampling methods
'Study' 카테고리의 다른 글
[Linear Algebra] 12 - 22 필기한 내용 (0) | 2024.02.27 |
---|---|
[Linear Algebra] 02 - 11 필기한 내용 (0) | 2024.02.27 |
[CNN] architecture (0) | 2024.02.06 |
[LG Aimers] Module 6. Deep Learning (1) | 2024.01.26 |
[CS231n] Assignment 1 - Two Layer Net (0) | 2024.01.16 |