Convolutional Neural Networks

Implement the foundational layers of CNNs (pooling, convolutions) and stack them properly in a deep network to solve multi-class image classification problems.

Learning Objectives

• Explain the convolution operation
• Apply two different types of pooling operations
• Identify the components used in a convolutional neural network (padding, stride, filter, ...) and their purpose
• Build a convolutional neural network
• Implement convolutional and pooling layers in numpy, including forward propagation
• Implement helper functions to use when implementing a TensorFlow model
• Create a mood classifer using the TF Keras Sequential API
• Build a ConvNet to identify sign language digits using the TF Keras Functional API
• Build and train a ConvNet in TensorFlow for a binary classification problem
• Build and train a ConvNet in TensorFlow for a multiclass classification problem
• Explain different use cases for the Sequential and Functional APIs

Computer Vision Problems

Vertical Edge Detection

Vertical Edget

Learning to detect edgets

Get Specific Edgets

Padding Image

Valid And Same Convolutions

Valid:

• nxn * fxf -> n-f+1 x n-f+1
• 6x6 * 3x3 -> 4x4

Strided Convolution

Strided Formula

Summary Of Convoultion

Technical Note On Cross-correlation vs. Convolution

Convolution In Math Textbook

Convolutions On RGB Images

Multiple Filters

Example Of A Layer

Exercise

Summary Of Convolution

Pooling Layer: Max pooling

Summary of MaxPooling

Number of Parameters

Here are the 5 typos:

1. 208 should be (553 + 1) * 8 = 608
2. 416 should be (558 + 1) * 16 = 3216
3. In the FC3, 48001 should be 400*120 + 120 = 48120, since the bias should have 120 parameters, not 1
4. Similarly, in the FC4, 10081 should be 120*84 + 84 (not 1) = 10164

(Here, the bias is for the fully connected layer. In fully connected layers, there will be one bias for each neuron, so the bias become In FC3 there were 120 neurons so 120 biases.)

5. Finally, in the softmax, 841 should be 84*10 + 10 = 850

Why Use Convolution

This is based on the equation:

Outline

LeNet-5

AlexNext

VGG-16

Residual block

Residual Network

The Problem of computational cost

1X1 Convolution

Inception Module

Inception Network

Origin Inception

https://knowyourmeme.com/memes/we-need-to-go-deeper

Mobile Net

Common Augmentation Method

1. Mirroring

2. Random Cropping

3. Color Shifting

Implementing Distortions During Training

Object Localization

Defining the target label y

Face verification vs. face recognition