Course Study with CSC321
[2021.04.03]
Topic : Lecture 13.3: Object Detection : SPPNET overview & details (Chanju, James)
Notes :
- https://drive.google.com/file/d/1IQHx6BPhMwCURbBuw6QBcBeLk8YSUnZg/view?usp=sharing [Chanju]
- https://drive.google.com/file/d/1bQAmtjUexd1lbpZgRB_sKC8VOwPMs064/view?usp=sharing [James]
Links :
SPP Net
- https://arxiv.org/pdf/1406.4729.pdf (original spp net thesis)
- https://driip.me/5743aed5-c630-4900-b367-9987a088661a (what is BoW approach in image?)
- https://n1094.tistory.com/30 (spp layer performance in classificaton & detection)
- https://blog.naver.com/laonple/220731472214 (Laon people, 내용 생략 심함)
- https://yeomko.tistory.com/14 (SPPnet 전반적 흐름 & 설명)
- https://www.youtube.com/watch?v=i0lkmULXwe0 (SPPnet 논문 강의 : 고려대학교 연구실)
- https://89douner.tistory.com/89 (SPPnet 보충 설명 , 자세한)
Fast R-CNN
- https://arxiv.org/pdf/1504.08083.pdf (original fast r-cnn thesis)
- https://fintecuriosity-11.tistory.com/73 (ablation study)
- https://yeomko.tistory.com/15 (how is end-to-end training possible?)
- https://deepsense.ai/region-of-interest-pooling-explained/ (spp vs roi pooling)
Covered through study
- SPP Net (overview) :
- What’s improved from R-CNN ? (idea, keywords)
- SPP Net flow (rough) (compared with R-CNN)
- SPP layer details (bin, BoW, how to calculate output)
- Practical training (Single-size, Multi-size training)
- Performance in fields (Classification, Detection)
- SPP Net Limits
- SPP Net (details) :
- how kM-d vector calculated through SPP layer?
- concepts could easily get confused (GAP, Inception module)
- what exact part of SPP Net is trainable?
- how spatial-pyramid-pooling acts different when training VS. test
SPP Net (overview)
SPP Net (details)
