Course Study with CSC321

[2021.05.08]

Topic : Lecture 13.5: Object Detection Faster R-CNN part1 (Chloe, James)

Notes :

• https://drive.google.com/file/d/1–1Wj2JcrLcxMHPWoJNo24sPg2TAa-Go/view?usp=sharing [Chloe]
• https://drive.google.com/file/d/16bjfABuBK1J-ejQRYNgOLjT2Z-iiqiu2/view?usp=sharing [James]

Links :

• https://arxiv.org/pdf/1506.01497.pdf (Faster R-CNN original thesis)
• https://www.youtube.com/watch?v=46SjJbUcO-c&t=1451s (기초개념 참고 유튜브 영상)
• https://deep-learning-study.tistory.com/464 (In/Out of RPN picture)
• https://herbwood.tistory.com/10 (Training RPN details, KR)
• https://www.telesens.co/2018/03/11/object-detection-and-classification-using-r-cnns/ (Training RPN details, EN)
• https://herbwood.tistory.com/11?category=867198 (코드로 이해하는 RPN)
• https://medipixel.github.io/post/2019-06-14-anchor-target/#ref_7 (코드로 이해하는 RPN loss)
• https://ganghee-lee.tistory.com/39 (FCN 참고자료 1)
• https://medium.com/hyunjulie/1%ED%8E%B8-semantic-segmentation-%EC%B2%AB%EA%B1%B8%EC%9D%8C-4180367ec9cb (FCN 참고자료 2)

Covered through study

Part 1

1. What’s improved? (or suggested?)

• 키워드별로 개념만, 뒤에 세부내용이 별도로 나옴

• RPN , region proposal networks ( kind of FCN?)

• Pyramids of images VS. Pyramids of filters VS. Pyramids of Anchors

1. Model architecture & Forward-pass (brief check)

• 마찬가지로 간단히

• how a single image passs through model

1. All about RPN

• 자세히
• Inputs & Outputs
• Anchor Box
  • what is Anchor box & what does translation-invariant means
  • how to refer anchor box to regression

Part 2

• Loss

  • what loss function is defined on RPN?

• Train

  • how to train RPN?

1. How RPN and Detector share feature maps?

• alternating training?

1. Implementation details

• 가능한 정도만
• used scales, anchor types

​

​

Faster R-CNN part 1

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​