Introduction to Convolutional Neural Networks With Image Classification Using PyTorch

  • Nemanja Milosevic

Your browser needs to be JavaScript capable to view this video

Try reloading this page, or reviewing your browser settings

You're watching a preview of subscription content. Log in to check access

In this video course, you will learn the basic principles of neural networks that are used to build models. You’ll start by seeing machine learning, neurons, activations, activation functions, weights, and how everything works under the hood. Next, you’ll cover the basics of the learning loop including how backpropagation and gradient descent work. Further, you will learn about convolutions, how they are inspired by the animal visual cortex, and how we use them in neural networks. One of the focuses of the course is image classification and detecting common objects in images. This has many uses in your day-to-day projects. We will be using the PyTorch open-source neural network library here.

The course will also cover current state-of-the-art neural network models and show how to use them even on smaller hardware. The video concludes by showing some common tricks with hyperparameter settings and regularization techniques, and how to use neural networks in production environments.

What You Will Learn

  • Discover the basics of neural networks and how they function

  • Work with convolutional neural networks

  • Use CNNs in your day-to-day work for image classification and other tasks

Who This Video Is For

Data scientists and machine learning and deep learning engineers.

In this video course, you will learn the basic principles of neural networks that are used to build models. You’ll start by seeing machine learning, neurons, activations, activation functions, weights, and how everything works under the hood. Next, you’ll cover the basics of the learning loop.

About The Author

Nemanja Milosevic

Nemanja Milosevic is a neural network researcher at UNSPMF, Serbia where he also teaches artificial intelligence among other courses. His primary research topic includes neural network robustness and susceptibility to adversarial attacks while he has also worked with image classification, NLP, time-series analysis, and many other machine learning tasks. He also delivered a talk on “Convolutional Neural Networks and Image Classification” at the recent EuroPython 2019 summit held at Basel, Switzerland.

 

Supporting material

View source code at GitHub.

About this video

Author(s)
Nemanja Milosevic
DOI
https://doi.org/10.1007/978-1-4842-5648-0
Online ISBN
978-1-4842-5648-0
Total duration
1 hr 20 min
Publisher
Apress
Copyright information
© Nemanja Milosevic 2020

Related content

Video Transcript

[MUSIC PLAYING]

Hello. Welcome to the course introduction to Convolutional Neural Networks and Image Classification. My name is Namanja Milosevic. And I come from University of Novi Sad, in Serbia, where I work as a teaching assistant and a machine learning researcher. And I will be your instructor on this course.

You can find my contact information here. This is my email, if you have any questions regarding the course or the materials. And this is also my Google Scholar link, if you are interested in my publications.

Let’s briefly go over this course agenda, what will be covered. In part 1, which is this part, we will go over some basic information about the course. We will talk about some prerequisites in knowledge. And we will also talk about some prerequisites regarding software that you will need, if you want to run the examples provided in this course.

In part 2 of the course, we will briefly introduce neural networks, as a machine learning algorithm. And we will talk about where they come in the ecosystem of artificial intelligence, deep learning, machine learning, and so on. We will also build our first neural network model in this part.

In the next part, part 3, we will cover what are convolutional neural networks, why do we need them, and how do they work, and why are they better than classic, traditional fully connected neural networks.

In part 4, the last part, we will briefly cover how can you publish your fully trained neural network models, so you can use them in your other applications. So, in other words, how you can expose the neural network APIs.

As far as course materials go, you have access to slides and videos. And all the code that is going to be used in this course will be on my GitHub, which you can find the link right here, on the screen.

About some course prerequisites and targeted audience for this course. So this course is for someone who is a software engineer who has some experience with programming or software engineering but is not familiar with machine learning or neural networks in general. You will have to know Python, the programming language, somewhat. We will be using Python 3.7. And some basic machine learning and neural network knowledge is needed. But its preferred, but it’s not necessary.

As far as software prerequisites go, you have two options. You can run everything locally. So all the examples you can run locally if you installed Pytorch, which is the tool from Facebook that we’ll be using for our experiments. The installation depends largely on your system, so based on your operating system, based on whether you have a graphics card with CUDA support.

So I think the best idea, if you want to run this locally, is to go to this link, https://pytorch.org/get started/locally/. And follow the wizard there, where you will get a command that you need to run to install everything.

The other options, which is more simple, is to use Google Colaboratory, which is a free service from Google, where you can run your Python code, and you also have some GPU, or graphic processing unit, support.