네이버 부스트캠프 모각공 캠페인 10일차 - CNN-Convolution은 무엇인가?

모든 설명 및 자료의 출처는 네이버 부스트코스의 <[부스트캠프  AI Tech 3기] Pre-Course>입니다.

(https://www.boostcourse.org/onlyboostcampaitech3/joinLectures/329424)

 

<인공지능 본격 탐구: 딥러닝 기초>

CNN-Convolution은 무엇인가?

 

Convolution

 

Continuous convolution

 

Discrete convolution

 

2D image convolution

 

Convolution

 

2D convolution

 

RGB Image Convolution

 

Stack of Convolution

 

Convolution 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), 3X3 Kernel

 

Exercise

What is the number of parameters of this model?

 

1X1 Convolution

Why?

Dimension reduction

To reduce the number of parameters while increasing the depth

e.g., bottleneck architecture

---

출처

[부스트캠프 AI Tech 3기] Pre-Course