Deep Learning From Scratch In Python

Understand Convolutional Neural Networks and Implement your Object-Detection Framework From Scratch

What you'll learn

Understand how Deep Neural Networks work, practically and mathematically

Understand Forward- and Backpropagation processes, mathematically and practically

Design and implement a Deep Neural Network for multi-class classification

Understand and implement the building blocks of Convolutional Neural Networks

Understand and Implement cutting-edge Optimization, Regularization and Initialization techniques

Train and validate a Convolutional Model on widely used datasets like MNIST and CIFAR-10

Understand and implement Transfer Learning

Use a Convolutional Model to create a Real-Time, Multi-Object Detection System

Requirements

No prior knowledge is required

Description

This course is for anyone willing to really understand how Convolutional Neural Networks (CNNs) work. Every component of CNNs is first presented and explained mathematically, and the implemented in Python.Interactive programming exercises, executable within the course webpage, allow to gradually build a complete Object-Detection Framework based on an optimized Convolutional Neural Network model. No prior knowledge is required: the dedicated sections about Python Programming Basics and Calculus for Deep Learning provide the necessary knowledge to follow the course and implement Convolutional Neural Networks.In this course, students will be introduced to one of the latest and most successful algorithms for real-time multiple object detection. Throughout the course, they will gain a comprehensive understanding of the Backpropagation process, both from a mathematical and programming perspective, allowing them to build a strong foundation in this essential aspect of neural network training.By the course's conclusion, students will have hands-on experience implementing a sophisticated convolutional neural network framework. This framework will incorporate cutting-edge optimization and regularization techniques, enabling them to tackle complex real-world object detection tasks effectively and achieve impressive performance results. This practical knowledge will empower students to advance their capabilities in the exciting field of Computer Vision and Deep Learning.

Overview

Section 1: Neural Networks Basics

Lecture 1 Introduction

Lecture 2 Intuition about Fully-Connected Networks

Lecture 3 Gradient Descent Algorithms

Lecture 4 Training, Validation and Testing

Section 2: Python Programming Basics

Lecture 5 Python for CNNs

Lecture 6 Working with Lists and Tuples

Lecture 7 Working with NumPy Arrays

Lecture 8 Object-Oriented Programming

Section 3: Calculus for Deep Learning

Lecture 9 The Derivative of a Function

Lecture 10 The Product, Quotient and Power Rules

Lecture 11 Derivatives by the Chain Rule

Section 4: Cost Functions and Backpropagation

Lecture 12 Backpropagation in Fully-Connected Networks

Lecture 13 The Softmax Activation Function

Lecture 14 The Cross-Entropy Cost Function

Lecture 15 Backpropagation in the Output Layer

Section 5: Building Blocks of Convolutional Neural Networks (CNNs)

Lecture 16 Introduction to Convolutional Networks

Lecture 17 Convolutions: Theory

Lecture 18 Convolutions: Implementing an Edge Detector

Lecture 19 Downsampling through Max Pooling

Section 6: Backpropagation in Convolutional Neural Networks

Lecture 20 Backpropagation in Convolutional Layers

Lecture 21 Backpropagation in Pooling Layers

Section 7: Integration of a Convolutional Model

Lecture 22 Defining a Convolutional Model

Lecture 23 The MNIST Dataset

Lecture 24 Filter Visualization

Section 8: Transfer Learning

Lecture 25 What is Transfer Learning

Section 9: Insights into Optimization and Regularization

Lecture 26 Fully-Convolutional Implementation

Lecture 27 The Vanishing Gradient and Dying ReLU Problems

Lecture 28 Parameters Initialization

Lecture 29 Learning Rate Decay

Lecture 30 The Adam Optimizer

Lecture 31 Testing the Optimized Model on MNIST

Lecture 32 Testing the Optimized Model on CIFAR-10

Section 10: Multiple Object Detection in Real-Time

Lecture 33 The YOLO Algorithm

Lecture 34 Testing the YOLO Algorithm

Everyone interested in really understanding Convolutional Neural Networks and willing to create their own Object Detection Framework in Python