티스토리 뷰
이전에 DetectoRS 이해를 위한 스텝을 나눴다.
2021/02/22 - [분류 전체보기] - Detecto RS를 이해하기 위한 기반 다지기
Detecto RS를 이해하기 위한 기반 다지기
DetectoRS가 너무너무 어렵기 때문에 하나하나 이해해 보기로 했다.
cho-akashic-records.tistory.com
우선, DetectoRS는 object detection 알고리즘이고, object detection 은 객체의 위치가 어디있는지 파악하는 Localization 과, 물체가 무엇인지 식별하는 classification 문제를 다룬다.
Deep learning에서의 Object Detection 은 크게 1-stage detector와 2-stage detector로 나눌 수 있다.
그림 1. 1-stage detector vs 2-stage detector 논문 흐름
1-stage detector, 2-stage detector의 차이와 전체 구조는 아래와 같다.
그림 2. 1-stage Detector의 전체 구조 (출처 : hoya012.github.io)
그림 3. 2-stage detector의 전체 구조(출처 : hoya012.github.io)
localization과 classification | Speed | Accuracy | example | |
1-satge detector | 동시에 진행 | 빠르다 | 낮다 | YOLO, SSD |
2-satge detector | 순차적으로 진행 | 느리다 | 높다 | RCNN, Fast RCNN, Faster RCNN |
여기까지가 결론이고 그 다음은 구성을 분석하여 원인을 알아본다.
+a) 보통 2-stage detector를 먼저 설명하는데, 그 이유는 주요 알고리즘 기준으로 2-stage가 먼저 나왔기 때문이다. 2-stage의 속도문제를 개선하기 위해서 나온 것이 1-stage이다.
출처 :
1. [Object Detection] one-stage detector vs two-stage detector jdselectron.tistory.com/101
2. [Object Detection] 1. Object Detection 논문 흐름 및 리뷰
3. Joseph Chet Redmon’s blog, “YOLO CVPR 2016”, https://goo.gl/Xj2Eik
4. 1-Stage detector와 2-Stage detector란? ganghee-lee.tistory.com/34
5. Object Detection sds.postech.ac.kr/wp-content/uploads/2020/07/object-detecion.pdf
6. 1-Stage detector 리뷰 chacha95.github.io/2020-02-26-Object-Detection3/
'AI > Baseline' 카테고리의 다른 글
Deformable Convolutional Networks(DCN) (0) | 2021.03.18 |
---|---|
1-stage detector 객체 검출법 (0) | 2021.03.01 |
Deformable Part Model(DPM) ; (0) | 2021.02.26 |
Selective Search, one of the Region Proposal (0) | 2021.02.22 |
Detecto RS를 이해하기 위한 기반 다지기 (1) | 2021.02.22 |
- Total
- Today
- Yesterday
- textual guidance
- yolov3 #tensorflow_convert
- ctfloader
- Arbitrary image stylization
- tensorflow convert
- /var/log/fontconfig.log
- StyleCLIP
- Deformable Part Model
- DetectoRS
- max-margin_loss
- 도메인 어댑테이션
- 전이학습
- /var/log/alternatives.log
- 첫논문
- alternatives.log
- 1-stage detection
- TransferLearning
- 샴네트워크
- neural network applications
- Source-free unsupervised domain adaptation
- Group Acitivty Recognition
- Chatgpt4
- torchmodel
- batch란
- /var/log/tallylog
- Deformable Convolutional Networks
- tf version porting
- /var/log/faillog
- 딥페이크탐지
- tf2to1
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 |