๐Ÿ”ด Lecture 09. | CNN Architectures

๋ฐฑ๊ฑดยท2022๋…„ 1์›” 21์ผ
0

Stanford University CS231n.ย 

๋ชฉ๋ก ๋ณด๊ธฐ
6/6

๋ณธ ๊ธ€์€ Hierachical Structure์˜ ๊ธ€์“ฐ๊ธฐ ๋ฐฉ์‹์œผ๋กœ, ๊ธ€์˜ ์ „์ฒด์ ์ธ ๋งฅ๋ฝ์„ ํŒŒ์•…ํ•˜๊ธฐ ์‰ฝ๋„๋ก ์ž‘์„ฑ๋˜์—ˆ์Šต๋‹ˆ๋‹ค.
๋˜ํ•œ ๋ณธ ๊ธ€์€ CSF(Curation Service for Facilitation)๋กœ ์ธ์šฉ๋œ(์ฐธ์กฐ๋œ) ๋ชจ๋“  ์ถœ์ฒ˜๋Š” ์ƒ๋žตํ•ฉ๋‹ˆ๋‹ค.

[์š”์•ฝ์ •๋ฆฌ]Stanford University CS231n. Lecture 09. | CNN Architectures

1. CONTENTS


1.1 Table

VelogLectureDescriptionVideoSlidePages
์ž‘์„ฑ์ค‘Lecture01Introduction to Convolutional Neural Networks for Visual Recognitionvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture02Image Classificationvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture03Loss Functions and Optimizationvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture04Introduction to Neural Networksvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture05Convolutional Neural Networksvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture06Training Neural Networks Ivideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture07Training Neural Networks IIvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture08Deep Learning Softwarevideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture09CNN Architecturesvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture10Recurrent Neural Networksvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture11Detection and Segmentationvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture12Visualizing and Understandingvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture13Generative Modelsvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture14Deep Reinforcement Learningvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture15Invited Talk: Song Han Efficient Methods and Hardware for Deep Learningvideoslidesubtitle
์ž‘์„ฑ์ค‘Lecture16Invited Talk: Ian Goodfellow Adversarial Examples and Adversarial Trainingvideoslidesubtitle

1.2 KeyWords

1.2.1 Convolusion

์ฐธ์กฐ : ํ•ฉ์„ฑ๊ณฑ ์—ฐ์‚ฐ๊ณผ ๊ต์ฐจ ์ƒ๊ด€ ์—ฐ์‚ฐ

  • ํ•ต์‹ฌ๊ฐœ๋…
    • ํ•ฉ์„ฑ๊ณฑ ์—ฐ์‚ฐ
    • ๋‘ ์‹ ํ˜ธ๋ฅผ ํ•ฉ์ณ ์ƒˆ๋กœ์šด ์‹ ํ˜ธ๋ฅผ ๋งŒ๋“œ๋Š” ์ˆ˜ํ•™์  ๋ฐฉ๋ฒ•
  • ์ข…์„ ์น  ๋•Œ ์ฝ˜๋ณผ๋ฃจ์…˜์„ ๋“ค๋ฆฌ๋Š” ์†Œ๋ฆฌ๋ผ๊ณ  ํ• ๋–„

[์ข…์„ ์น ๋•Œ ์ข… ์ž์ฒด๊ฐ€ ๋‚ด๋Š” ์†Œ๋ฆฌ] * [์ข…์†Œ๋ฆฌ๊ฐ€ ์ข… ๋‚ด๋ถ€ ๊ณต๊ธฐ์˜ ์ง„๋™์‹œ์ผœ ์šธ๋ฆฌ๋Š” ์†Œ๋ฆฌ]
= ๊ณต๋ช…์œผ๋กœ ๋“ค๋ฆฌ๋Š” ์†Œ๋ฆฌ

<ํ™•์žฅ> - ์—ฐ์†์œผ๋กœ ์น ๋•Œ
์ƒˆ๋กœ ์นœ ์†Œ๋ฆฌ * ์•ž์„  ์†Œ๋ฆฌ์˜ ์ž”ํ–ฅ
์ž…๋ ฅ + ๋ฉ”๋ชจ๋ฆฌ => ์ถœ๋ ฅ

1.2.1 ์ˆ˜์šฉ์˜์—ญ

์ฐธ์กฐ :

  • ํ•ต์‹ฌ๊ฐœ๋…

    • ๊ณ„์ธต์ด ๊นŠ์–ด์งˆ ์ˆ˜๋ก ํ•„ํ„ฐ-1 ๋งŒํผ ์ˆ˜์šฉ์˜์—ญ์ด ์ปค์ง
    • O : ์ถœ๋ ฅ / F : ์ปจ๋ณผ๋ฃจ์…˜ ํ•„ํ„ฐ / N : ์ž…๋ ฅ / S : ์Šคํ”„๋ผ์ด๋“œ / P : ํŒจ๋”ฉ

    O=(Nโˆ’2P)S+1O=\frac{(N-2{P})}{S} + 1

2. Flow

2.1 title

2.1.1 sub-title

2.2 title

2.3 title

3. Summary

3.1 OverView


Slide
  • ์ง€๋‚œ ์‹œ๊ฐ„ ์š”์•ฝ ํŠน๋ณ„ํ•œ ๋‚ด์šฉ์—†์Œ



3.2 LeNet-5


Slide
  • ์šฐํŽธ๋ฌผ์— ํ•„๊ธฐ์ฒด๋กœ ์“ฐ์ธ ์šฐํŽธ๋ฒˆํ˜ธ๋ฅผ ์ธ์‹ํ•˜๋Š” ์ตœ์ดˆ์˜ ์ปจ๋ฒŒ๋ฃจ์…˜
  • ํ™œ์„ฑํ•จ์ˆ˜๋Š” ์‹œ๊ทธ๋ชจ์ด๋“œ ๋˜๋Š” ํ•˜์ดํผ๋ณผ๋ฆญ ํƒ„์  ํŠธ๋ฅผ ์‚ฌ์šฉ
  • ํ’€๋ง ๊ณ„์ธต์€ ์•กํ‹ฐ๋ฒ ์ด์…˜ ๋งต์˜ ํฌ๊ธฐ๋ฅผ ์ ˆ๋ฐ˜์œผ๋กœ ์ค„์ž„
  • ํŒŒ๋ผ๋ฏธํ„ฐ ์ˆ˜๊ฐ€ 60K๋กœ ์ž‘์€ ๋ชจ๋ธ
  • ์ปจ๋ฒŒ๋ฃจ์…˜ ๋ ˆ์ด์–ด์™€ ํ’€๋ง ๋ ˆ์ด์–ด๋Š” ์‹œ๊ฐ ์˜์—ญ์„ ๋ชจ๋ธ๋ง
  • FC๋Š” ์—ฐ๊ด€ ์˜์—ญ์„ ๋ชจ๋ธ๋ง



3.3 AlexNet


Slide
  • 2๊ฐœ์˜ GPU (๋‹น์‹œ GTX580 GPU)
  • ์ปจ๋ฒŒ๋ฃจ์…˜ ํ•„ํ„ฐ๋ฅผ ๋‘๊ทธ๋ฃน์œผ๋กœ ๋‚˜๋ˆˆ ๋’ค ๊ทธ๋ฃน๋ณ„๋กœ GPU ํ• ๋‹นํ•˜์—ฌ ์ฒ˜๋ฆฌ
  • ์ค‘๊ฐ„ ๊ณ„์ธต์—์„œ ์ •๋ณด๋ฅผ ๊ตํ™˜
  • ์ตœ์ข…์€ ํ•œ์ชฝ์œผ๋กœ ํ•ฉ์นจ
  • 227x227xx3์˜ ์ž…๋ ฅ๋ฐ์ดํ„ฐ
  • 48 convolution Filter



3.4 VGGNet


SlideTitleref
--



3.5 GoogLeNet


SlideTitleref
--



3.6 ResNet


SlideTitleref
--



3.7 Compare


SlideTitleref
--



3.8


SlideTitleref
--



3.9


SlideTitleref
--



3.10


SlideTitleref
--



Reference

profile
๋งˆ์ผ€ํŒ…์„ ์œ„ํ•œ ์ธ๊ณต์ง€๋Šฅ ์„ค๊ณ„์™€ ์Šคํƒ€ํŠธ์—… Log

0๊ฐœ์˜ ๋Œ“๊ธ€