Can computers see the world as humans do? While we still don’t understand how human vision functions to be able to answer that question, the advent of more powerful artificial intelligence has catapulted computer vision – the field of study that tries to understand and replicate parts of the human vision system – to revolutionizing a range of industries as well as having an impact on our everyday lives. As a result, there is a large demand for engineers in computer vision and it is increasingly important also for non-specialists to have an understanding of this area.

In this article, we will look at how you can best approach mastering the area of computer vision, depending on your starting point and your goals.

We present our main recommendations for courses in greater detail below. Note that at the end of this review, you will also find a more comprehensive list of good alternative courses that, while they did not make our short list of recommended courses, may be a better fit for your specific learning needs.

Introduction

The field of computer vision has its roots in the early work on artificial intelligence at universities in the 1960s. Alongside the development of computers, computer vision has also been improving gradually. One of the greatest practical breakthroughs that many of us are familiar with was optical character recognition (OCR) in the 1970s – the possibility for computers to read text.

But even if computer vision had already come a long way through “conventional” computing, it has received much greater attention lately through the advancement of deep learning and neural networks, helping shift the focus to object recognition and its practical applications. In our current world, with vast amounts of visual data produced and shared daily (more than 3 billion images per day and 720,000 hours of video), improved artificial intelligence techniques that can be trained on this flow of information and the computing power needed for processing it, computer vision has become more powerful and accessible.

The applications for computer vision have grown exponentially as a result, with the software and hardware computer vision market expected to reach $48.6 billion in 2022. This will help transform several aspects of our economy and everyday life, from manufacturing processes via augmented reality and video surveillance to self-driving cars. To design and implement these new applications, there is a large and growing need for computer vision engineers – employers have increased their demand for AI-related roles by over 100% since 2019 – with salaries to match.

Study paths: How do I choose my online course?

As for any area of study, how best to approach this field depends on what your objectives are. For computer vision, these tend to fall in the following broad categories:

Aiming to specialize in computer vision: If you are an aspiring computer vision developer, you have an exciting path ahead of you, with several good options.
- Follow our overall recommended course or our fundamentals recommendation, depending on your starting point and objectives. If you are not able to afford these options, follow our recommended free course.
Applying computer vision tools: Developers with other specializations can do a lot with modern computer vision tools that does not require them to have a deep understanding of the area.
- Follow our recommended crash course, introductory lab, and course with the best coverage of different tools.
Strategic overview: Many will benefit from understanding the applications and implications of computer vision better, without necessarily looking to program anything yourself.
- Follow our recommended high-level course.
Already a specialist: Those who are already working in computer vision will benefit from continuing education.
- Follow our recommended course for experts or go through our full list to find the course that best suits your specific learning needs.

While some universities offer good face-to-face programs that cover computer vision, there is luckily also a wide range of online courses that you can complete remotely, while having a similarly high quality and allowing you to mix and match courses as needed to meet your objectives. We have taken a comprehensive look at the available online courses and have tested them out to provide our recommendations to aspiring students.

Prerequisites: What do I need to know before getting started?

As computer vision in its modern applications combines methods for image processing with artificial intelligence techniques, it is not a topic that is suitable for those who are complete beginners to programming, data science, statistics, and artificial intelligence. As they would otherwise be too big and have a too wide scope, most courses on the topic have prerequisite knowledge in these areas. While most of the courses on this list will not control that you have these prerequisites, it is strongly recommended that you make sure you have the following before moving forward, so that you can digest and learn from the course material:

Programming: Python is by far the most common programming language used in courses on computer vision, with intermediate programming skills recommended or required. In our list of best Python courses, our overall recommendation is Python for Everybody from Coursera (which a dedicated student can complete in a lot less time than the estimated 8 months). If you need a more intense and project-focused program with mentoring, we recommend one of the nanodegrees from Udacity – Programming for Data Science with Python or AI Programming with Python if you are completely new to Python, or the Intermediate Python Nanodegree if you already have basic familiarity with the language.
Math: You need knowledge in algebra and statistics in several data science fields, and it is particularly important in areas using artificial intelligence. If you need to brush up on your math skills, our recommendation is Brilliant – they have courses in a very engaging format on the topics you would be expected to be up to speed on, including algebra, statistics (including probability and regression), and calculus.
Artificial intelligence (AI): While some of the courses on this list will use computer vision as the entry point into learning about artificial intelligence, but most of them expect you to already have a grasp of the main concepts and how they are implemented. Check out our lists of recommended courses for artificial intelligence, machine learning, and deep learning. Our main recommendation for anyone looking to get into computer vision would be the AI Programming with Python Nanodegree as it is a very high-quality guided program which will also give you the solid foundation in Python that you will need if specializing in this field.

Exactly what is needed will of course depend on the specific course and a couple of the less programming-focused courses on this list have no prerequisites.

Recommended courses

Best Overall Course: Computer Vision Expert Nanodegree (Udacity x Affectiva & Nvidia)

Pros

High quality, practice-oriented curriculum designed together with industry leader
Excellent instructors
Includes dedicated guidance and support for learners

Cons

For intermediate to advanced programmers – not the best option if you do not have the prerequisite experience

With its 3-month curriculum focusing on teaching those with previous programming and machine learning experience, the Computer Vision Expert Nanodegree on Udacity is an intensive course that gives students expertise to become fully-fledged computer vision engineers. Designed together with Nvidia – an industry leader in computer vision – this course comes with Udacity’s dependably high-quality delivery, project-focused curriculum, mentoring, and career support services, and is our overall pick for anyone looking to start or upgrade their career in the area of computer vision.

Note that as this is an advanced topic, there are prerequisites, like for most courses on our list. Students are expected to have the following:

Intermediate to advanced Python experience.
Intermediate statistics background, including probability
Intermediate knowledge of machine learning techniques, including familiarity with backpropagation and examples of neural network architecture
Familiarity with a deep learning framework like TensorFlow, Keras, or PyTorch.

The program is divided into three component courses, each focusing on a portfolio-building project:

Introduction to Computer Vision – focuses on the fundamentals of computer vision, but already at this stage applying deep learning techniques with a project on facial keypoint detection.
Advanced Computer Vision and Deep Learning – takes learners through more advanced convoluted neural networks (CNNs) architecture, recurrent neural networks (RNNs), and attention mechanisms, and brings it all together in a project on image captioning combining CNNs and RNNs.
Object Tracking and Localization – turns more to applications for mapping the environment with optical flow, feature matching, localization, and graph slam, to help you build a project on the localization techniques used in autonomous vehicle navigation.

The program has been around since 2018, but has been kept up-to-date on points where the technology has moved forward significantly.

As the course has a strong focus on AI applications of computer vision, it is not the best option for learners who want to start out with learning about the theoretical fundamentals of computer vision outside of AI. If this is important to you, we instead recommend first following the First Principles of Computer Vision Specialization from Coursera and Columbia University.

But if a career in computer vision is what you’re looking for, and you already have the necessary prerequisites, then this should be an easy choice. There are not any alternatives that we know of that will give you even nearly the degree of high-quality guidance and support in this particular field. Even if you don’t have everything yet, you may still want to aim to take this course after brushing up your skills with our recommendations for this in the intro. Your main option otherwise for a similarly high-quality offering is to go for a wider program in artificial intelligence, machine learning, or deep learning that also covers computer vision.

The cost of Udacity Nanodegrees, at $399/month or with a 15% discount if paying up front for the estimated duration of the program, is higher than most other more “self-service” style courses. But this reflects both the degree of work that has gone into the program design, and, more importantly, the support you are getting from teachers and assistants.

Best for Fundamentals: First Principles of Computer Vision Specialization (Coursera x Columbia University)

Time-limited offer
$100 USD off your first year of Coursera Plus Annual (expires 1 April 2024)

Pros

Comprehensive introduction to all "conventional" aspects of computer vision
Good starting point before continuing with applying AI to computer vision

Cons

Although listed as beginner-level, difficult without existing familiarity with computer vision

How does human vision work and how does it relate to how a digital camera functions? If you are interested in getting a solid understanding of the fundamentals of computer vision before learning more about how artificial intelligence can be applied to it, the First Principles of Computer Vision Specialization from Coursera and Columbia University is an excellent choice.

Note that while this course is categorized by Coursera as beginner level, if you have no prior familiarity with computer vision, you will likely struggle with the material, especially in the first course which moves at a rapid pace. We have changed the categorization to intermediate in our list, but as long as you are prepared to take extra time with the first course and pause frequently to research the topics mentioned, this course should be manageable as long as you have the prerequisites in math (fundamentals of linear algebra and calculus).

The course only touches on artificial intelligence applications at the end of the final course – looking at neural networks for object recognition. This makes for a great pivot point to our overall recommended course or another course on the list that focuses more on AI in computer vision.

Best Free Course: Introduction to Computer Vision (Udacity x Georgia Tech)

Pros

Excellent coverage of fundamentals of computer vision
Although free, high quality course material and good instructors

Cons

No personalized support

Introduction to Computer Vision is a solid free course offering developed by two leading educational actors in this field: Udacity (which also provides our overall top pick in this list) together with Georgia Tech for self-driven learners with existing intermediate math and programming skills.

With the focus being on understanding and applying low and mid-level algorithms to analyze images, the course provides a great introduction to most of the traditional computer vision techniques. However, it does not cover machine learning to any greater extent.

Even though the subject matter is advanced, the course instructors present it clearly. Although not necessarily to everyone’s liking, they do try to keep things a bit more lively by injecting some humor into their delivery.

While it has a well-designed curriculum, this course is primarily designed for those studying on a budget – it does not have the same level of interactivity and support as Udacity’s Nanodegrees. This is a challenging, graduate-level course – if you would like to increase your chances of succeeding in this area, we strongly recommend Udacity’s full Nanodegree instead. On the other hand, it does not cost you anything but your time to test this one out.

Best High-Level Overview: Computer Vision - Image Understanding for Efficient Business and Industry (FutureLearn x Lulea University of Technology)

Pros

Useful for programmers and non-programmers alike
Provides a higher-level perspective missing from the other courses on the list

Cons

Not very practically oriented

Are you interested in the applications of computer vision but you’re not a programmer? Computer vision is a very powerful tool and has the potential to transform several aspects of business – as such, it’s important for executives, managers, programmers in other fields, and many others to understand its possibilities and implications.

Designed for those interested in the topic without wanting to get straight into the programming implementation of it, the Computer Vision – Image Understanding for Efficient Business and Industry course from FutureLearn and Luleå University of Technology will take you through a high-level overview of the concepts and applications of computer vision. At an introductory level without any prerequisites, the four hours per week for three weeks is manageable for most.

Best Crash Course: Introduction to Deep Learning with OpenCV (LinkedIn Learning)

Pros

Quick overview of the main concepts to help orient yourself

It can be difficult to know where to start if you are not already familiar with computer vision. For those who already have some programming experience, the Introduction to Deep Learning with OpenCV from LinkedIn Learning is a good option for a first step. It provides a brief introduction to OpenCV and will give you a decent overview of some of the main frameworks and concepts, without getting into any major practical applications or projects.

As the course clocks in at only 49 minutes, this clearly will not let you master computer vision, but Jonathan Fernandes, the course instructor, does a fine job in helping you take your first few steps in building your understanding of the field.

Best Introductory Lab: Computer Vision 101 - Let's Build a Face Swapper in Python (Skillshare)

Special offers from Skillshare:
Get 30% off of a Premium Skillshare Membership
Test Skillshare for free with no-questions-asked cancellation: 30-day free trial

Pros

Let's you get your hands dirty pretty much instantly

Cons

More inspirational than providing much in terms of lasting skills

Sometimes you just want to get your hands dirty without spending too much time learning about the theory behind something. If you already have experience with Python and want a quick practical introduction to what you can do with computer vision, then Computer Vision 101 – Let’s Build a Face Swapper in Python is for you. This practically oriented short course from Skillshare will quickly get you set up with OpenCV and have you running your own example project in no time. Good production value and delivery by Alvin Won.

Best Tool Coverage: Modern Computer Vision - PyTorch, Tensorflow2 Keras & OpenCV4 (Udemy)

Offers from Udemy:
New customer offer! Top courses from $14.99 when you first visit Udemy (expires 31 March 2024).

Pros

Good option for getting an overview of the capabilities of multiple tools for those who prefer teaching via explained code snippets

Cons

Lacking contextual and theoretical explanations

There are several tools and libraries that are relevant for computer vision programmers. If you are a programmer looking to get introduced to some of these, we recommend the Modern Computer Vision course from Rajeev Ratan at Udemy. This course covers a wide range of tools with PyTorch, TensorFlow2 and OpenCV4 with a focus on explaining code snippets.

The course instructor is not the most experienced expert, and his explanations are lacking in explanations of overall principles and theory. However, he does a fine job introducing you to the practical applications of the tools covered and if you like learning via code snippet explanations, this course is a good option for you.

Best for Microsoft Azure: Applied Artificial Intelligence - Computer Vision and Image Analysis (FutureLearn x CloudSwyft)

Pros

Excellent course for those working in Microsoft Azure
Accredited by Microsoft

The Applied Artificial Intelligence – Computer Vision and Image Analysis course from FutureLearn and CloudSwyft is an excellent introduction for computer vision using this platform, including both Microsoft tools and OpenCV. The course instructors take you through classical image analysis techniques before turning to deep learning object classification and applying the Microsoft ResNet Convolutional Neural Network (CNN) and the Microsoft Cognitive Toolkit.

The course is also accredited by Microsoft, helping ensure that you are getting the most relevant skills for working in computer vision using their tools. Note that the course is part of an ExpertTrack on AI in Microsoft Azure, and if you know you will be working on this platform, you would be well-advised to also take the other three related courses.

Best for MATLAB: Introduction to Data, Signal, and Image Analysis with MATLAB (Coursera x Vanderbilt University)

Time-limited offer
$100 USD off your first year of Coursera Plus Annual (expires 1 April 2024)

Pros

Excellent course for those working in MATLAB
Clear instruction, if a little bit long-winded

If you’re one of the millions of engineers and scientists using MATLAB in your work, you may already be aware that this programming language is very capable in the area of computer vision. The Introduction to Data, Signal, and Image Analysis with MATLAB from Coursera and Vanderbilt University is a very good course introducing those who already have intermediate MATLAB knowledge to its data, signal, and image processing functionalities.

The course features useful assignments. The instructor, Jack Noble, has put a lot of effort into making sure the instructional videos are comprehensive and pedagogical – but they are a bit on the long side for some learners.

This course is part of the MATLAB Programming for Engineers and Scientists Specialization.

Best for Experts: Advanced Computer Vision with TensorFlow (Coursera x DeepLearning.AI)

Time-limited offer
$100 USD off your first year of Coursera Plus Annual (expires 1 April 2024)

Pros

Excellent course for those working in TensorFlow
Great instructors
Curriculum focused on practical applications

Cons

Advanced level and high-paced without much support

TensorFlow is one of the most in-demand and popular open-source deep learning frameworks available today and is used for cutting-edge applications of computer vision. The practically oriented Advanced Computer Vision with TensorFlow from Coursera and Deeplearning.AI will help programmers deepen their knowledge of how to use this framework through a set of projects on object detection, image segmentation, and visualization and interpretability.

Designed for early and mid-career software and machine learning engineers with a foundational understanding of TensorFlow, it is not recommended as a standalone course for most students, but can be combined with the other advanced courses in the TensorFlow Advanced Techniques Specialization offered on Coursera. If you need an introduction to TensorFlow, the DeepLearning.AI TensorFlow Developer Professional Certificate is your best starting point.

The 10 Best Online Courses for Computer Vision

Computer Vision - Table of Contents

Introduction

Study paths: How do I choose my online course?

Prerequisites: What do I need to know before getting started?

Recommended courses

Best Overall Course: Computer Vision Expert Nanodegree (Udacity x Affectiva & Nvidia)

Cons

Best for Fundamentals: First Principles of Computer Vision Specialization (Coursera x Columbia University)

Cons

Best Free Course: Introduction to Computer Vision (Udacity x Georgia Tech)

Cons

Best High-Level Overview: Computer Vision - Image Understanding for Efficient Business and Industry (FutureLearn x Lulea University of Technology)

Cons

Best Crash Course: Introduction to Deep Learning with OpenCV (LinkedIn Learning)

Best Introductory Lab: Computer Vision 101 - Let's Build a Face Swapper in Python (Skillshare)

Cons

Best Tool Coverage: Modern Computer Vision - PyTorch, Tensorflow2 Keras & OpenCV4 (Udemy)

Cons

Best for Microsoft Azure: Applied Artificial Intelligence - Computer Vision and Image Analysis (FutureLearn x CloudSwyft)

Best for MATLAB: Introduction to Data, Signal, and Image Analysis with MATLAB (Coursera x Vanderbilt University)

Best for Experts: Advanced Computer Vision with TensorFlow (Coursera x DeepLearning.AI)

Cons

Markus Forsberg

Read next:

Es Devlin’s MasterClass: Turn Ideas into Art and Unleash Your Creative Potential

Focused Question Clusters: Learning through Questions

The Best SAT Prep Books for Reading, Writing, and Math