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 |
Tags
- 자료구조
- re-identification
- 3D
- OpenCV
- Object Tracking
- attention
- center loss
- Computer Vision
- NLP
- cv2
- 알고리즘
- transformer
- Knowledge Distillation
- deep learning
- 딥러닝
- Python
- 파이썬
- 논문 구현
- 임계처리
- 프로그래머스
- Threshold
- Deeplearning
- numpy
- flame
- point cloud
- reconstruction
- level2
- 스택
- Object Detection
- 큐
Archives
- Today
- Total
목록컴퓨터비전 (1)
공돌이 공룡의 서재
![](http://i1.daumcdn.net/thumb/C150x150/?fname=https://blog.kakaocdn.net/dn/b41Nn9/btrcOek6knn/o95VJiuIXfCzncqkA4FQm0/img.png)
Swin Transformer: Hierarchical Vision Transformer using Shifted Windows https://arxiv.org/abs/2103.14030 [1] Introduction Transformer는 NLP에서는 좋은 성능을 갖지만, Visual domain에 적용할 때 왜 어려운가?로 시작하고 있다. 이에 대한 원인으로 2가지를 제시하고 있다. scale : NLP에서는 word token을 사용하고, 크기가 고정적이지만, visual element는 스케일에 따라 다양하다. 다양하게 될 때 문제는 attention을 적용할 때 문제가 생긴다는 점이다. 기존 transformer 기반 모델들은 고정적인 크기로 가정하기 때문이다. high resolution : ..
딥러닝/Vision
2021. 8. 21. 03:24