티스토리 뷰
(1)
2024.03.25 - [AI] - 0325 papers + CI-UDA
(2)
공식까지 빠지지 않고 처음부터 끝까지 다읽은 논문은 처음이다. 그런 의미에서 논문에 있는 말을 번역한 것이 아닌 내가 이해한 바를 따로 정리해 보려고 한다.
class-level 에서는 confident 한 target sample과 src domain 간의 feature representation 을 맞추고(align), instance-level에서는 unconfident target sample의 robust featrue representation을 활용하는 Contrastive Algignment를 제안한다. 이를 통해 domain adaptation의 performance을 향상시키고자 한다.
- 여기서 class-level과 instance-level이 헷갈렸는데 그림으로 치면 이런 것 같다.

그런데 시간이 부족해 일단 중요단어 10개만 뽑아보자.
ProCA - memory bank
biased source information
negative trnsfer
Continual Domain Adaptation
catatrophic forgetting
contrastive learning - contrastive alignment - adversarial learning(feature extractor VS discriminator)
Prototypes(class-level prototype)
domain-invarint knowlege
pseudo Label distillation
feature centroid
'AI > 주워들은 것들' 카테고리의 다른 글
torch)메모리 부족 이슈로 Data loader가 뻑사리 날 때 insufficient shared memory (1) | 2024.10.23 |
---|---|
Triplet loss란? (2) | 2024.04.15 |
soft label vs pseudo label (0) | 2024.03.25 |
Domain Gereralization과 Domain Adaptation 차이 (0) | 2024.03.23 |
Fisher Discriminant Analysis(선형판별분석) (4) | 2024.03.23 |
공지사항
최근에 올라온 글
최근에 달린 댓글
- Total
- Today
- Yesterday
링크
TAG
- 샴네트워크
- /var/log/alternatives.log
- StyleCLIP
- yolov3 #tensorflow_convert
- TransferLearning
- Arbitrary image stylization
- batch란
- Deformable Convolutional Networks
- neural network applications
- alternatives.log
- ctfloader
- max-margin_loss
- tensorflow convert
- /var/log/faillog
- textual guidance
- DetectoRS
- /var/log/fontconfig.log
- Source-free unsupervised domain adaptation
- 첫논문
- 딥페이크탐지
- 1-stage detection
- Chatgpt4
- /var/log/tallylog
- tf2to1
- Group Acitivty Recognition
- tf version porting
- Deformable Part Model
- 도메인 어댑테이션
- 전이학습
- torchmodel
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 | 31 |
글 보관함