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
- Deeplearning
- NLP
- numpy
- 임계처리
- attention
- Object Tracking
- Python
- 큐
- reconstruction
- 파이썬
- transformer
- Object Detection
- Computer Vision
- 스택
- point cloud
- cv2
- flame
- level2
- re-identification
- Threshold
- 자료구조
- 3D
- center loss
- 논문 구현
- OpenCV
- 프로그래머스
- Knowledge Distillation
- 딥러닝
- deep learning
- 알고리즘
Archives
- Today
- Total
목록byte (1)
공돌이 공룡의 서재

Introduction 기존의 SOTA MOT 방법들은 low confidence detection box를 제거하기 위해 detection box 로 인한 true positive / false positive 문제를 다뤘다. →threshold를 기준으로 detection box 를 제거하다보니 생기는 true positive / false positive 문제를 다뤄야만 했다. FairMOT만 봐도, 0.5 가 넘는 box만 선택해서 사용한다. 이 논문의 요지는 과연 기존 방식대로 low confidence detection box를 지우는 것은 옳은 접근일까? 에서 시작된다. 저자들은 그렇지 않다라는 입장에서 문제에 접근하고 있다. → low confidence detection box도 obje..
딥러닝/Tracking
2022. 4. 12. 00:43