Lecture 16: Convolutional Neural Networks (II)

Content locked

The catalog is public. Reading this content requires approved access.

Introduction

Review: 2-D Convolution

Basic Example

Multiple Input Channels

Why Multiple Channels?

Filter Terminology

Example Calculation

Output Shape Formula

Example Problem

Pooling Layer

2-D Max Pooling

Average Pooling

CNN Architecture Intuition

Feature Learning Hierarchy

Classification Example: Cats vs. Dogs

Evolution of CNN Architectures

Timeline Overview

LeNet Architecture (1989)

Architecture Details

MNIST Dataset

LeNet in PyTorch

AlexNet Architecture (2012)

Key Improvements Over LeNet

Architecture Layers

Key Innovations

Parameter Count Example

VGG Architecture (2014)

VGG Block Structure

Key Insight

The Depth Problem

Why Not Just Add More Layers?

The Identity Insight

ResNet Architecture (2015)

Residual Connections

Residual Block Structure

ResNet Configurations

Benefits of Residual Connections

ImageNet Progress

Practice Problems

Key Takeaways