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 | 31 |
Tags
- 프로그래머스
- DFS
- Coding Test
- BFS
- 자연어처리
- 백준
- 탐욕법
- retinaNet
- 클래스 불균형
- 논문리뷰
- 코딩 테스트
- AlexNet
- 딥러닝
- dl
- Deep Learning
- VGGNet
- 그리디
- GoogleNet
- ResNet
- greedy
- Vision Transformer
- GPT
- Attention is all you need
- algorithm
- 알고리즘
- vit
- LeNet
- Transformer
- Focal loss
- greedy algorithm
Archives
- Today
- Total
목록retinaNet (1)
자율주행 미래를 위한 대학원생

이번에 리뷰할 논문은 'Focal Loss for Dense Object Detection' 이다. 이 논문은 RetinaNet이라는 모델을 제안하였다. 그리고 기존의 Object Detector들은 Class Imbalance 문제를 해결하기 어려웠다. Class Imbalance란 이미지 내에서 객체의 영역이 배경 영역보다 훨씬 적어 발생하는 클래스 불균형 문제이다. 이러한 easy negative는 학습이 비효율적이고, 모델의 성능을 저하시키게 된다. two-stage detector에서는 이 문제를 region proposal과 sampling heuristic을 통해 해결하지만 one-stage detector에는 적용이 불가하여 고안된 것이 본 논문에서 제시하는 Focal Loss 이다. Fo..
Paper Riview
2024. 1. 3. 13:43