[부스트캠프 AI Tech 5기] Pre-Course : (25) Convolutional Neural Networks

📖 Convolution

$O_{11} = I_{11}K_{11} + I_{12}K_{12} + I_{13}K_{13} + I_{21}K_{21} + I_{22}K_{22} + I_{23}K_{23} + I_{31}K_{31} + I_{32}K_{32} + I_{33}K_{33} + bias$

$O_{12} = I_{12}K_{11} + I_{13}K_{12} + I_{14}K_{13} + I_{22}K_{21} + I_{23}K_{22} + I_{24}K_{23} + I_{32}K_{31} + I_{33}K_{32} + I_{34}K_{33} + bias$

$O_{13} = I_{13}K_{11} + I_{14}K_{12} + I_{15}K_{13} + I_{23}K_{21} + I_{24}K_{22} + I_{25}K_{23} + I_{33}K_{31} + I_{34}K_{32} + I_{35}K_{33} + bias$

$O_{14} = I_{14}K_{11} + I_{15}K_{12} + I_{16}K_{13} + I_{24}K_{21} + I_{25}K_{22} + I_{26}K_{23} + I_{34}K_{31} + I_{35}K_{32} + I_{36}K_{33} + bias$

• 2D convolution in action
  

📖 RGB Image Convolution

📖 Stack of Convolutions

📖 Convolutional Neural Networks

• CNN consists of convolution layer, pooling layer, and fully connected layer.
  • Convolution and pooling layers: feature extraction
  • Fully connected layer: decision making (e.g., classification)
    

📖 Convolution Arithmetic (of GoogLeNet)

📖 Stride

📖 Padding

📖 Stride? Padding?

📖 Convolution Arithmetic

• Padding (1), Stride (1), 3 × 3 Kernel
  

• What is the number of parameters of this model?

  • The answer is $3 \times 3 \times 128 \times 64 = 73,728$

📖 Exercise

• What is the number of parameters of this model?

• 1) $11 \times 11 \times 3 \times 48 \times 2 \approx 35k$
• 2) $5 \times 5 \times 48 \times 128 \times 2 \approx 307k$
• 3) $3 \times 3 \times 128 \times 2 \times 192 \times 2 \approx 884k$
• 4) $3 \times 3 \times 192 \times 192 \times 2 \approx 663k$
• 5) $3 \times 3 \times 192 \times 128 \times 2 \approx 442k$
• 6) $13 \times 13 \times 128 \times 2 \times 2048 \times 2 \approx 177M$
• 7) $2048 \times 2 \times 2048 \times 2 \approx 16M$
• 8) $2048 \times 2 \times 1000 \approx 4M$

📖 1x1 Convolution

• Why?
  • Dimension reduction
  • To reduce the number of parameters while increasing the depth
  • e.g., bottleneck architecture

<이 게시물은 최성준 교수님의 'Convolutional Neural Networks' 강의 자료를 참고하여 작성되었습니다.>

본 포스트의 학습 내용은 [부스트캠프 AI Tech 5기] Pre-Course 강의 내용을 바탕으로 작성되었습니다.
부스트캠프 AI Tech 5기 Pre-Course는 일정 기간 동안에만 운영되는 강의이며,
AI 관련 강의를 학습하고자 하시는 분들은 부스트코스 AI 강좌에서 기간 제한 없이 학습하실 수 있습니다.
(https://www.boostcourse.org/)