(23.01.22)
convolution neural network 사용할 때의 hyperparameter
1) convolution
- filter 크기
- filter 개수
2) pooling
- window 크기
- window stride
3) fully-connected
- layer의 개수
- neurons의 개수
neural network 아키텍쳐 종류
- one to one: vanila nerual networks
- one to many: image captioning (image -> sequence of words)
- many to one: sentiment classification (sequence of words -> sentiment(감정))
- many to many(1): machine translation (sequence of words -> sequence of words)
- many to many(2): video classification on frame level
728x90
'Computer Vision > 메모 + 정리' 카테고리의 다른 글
[Depth Estimation] Depth Estimation 개론 (0) | 2023.04.30 |
---|---|
[WSSS] Weakly-supervised semantic segmentation (0) | 2023.02.14 |
[Segmentation] Seed Region Growing algorithms (0) | 2023.02.12 |
IoU & dice score (0) | 2023.01.25 |
[CS231n] Image classification - nearest neighbors & linear classification (0) | 2022.07.05 |