Learning Path for Computer Vision Engineers

Click on the above“Beginner Learning Vision” to choose tostar or “pin”

Essential knowledge, delivered at the first time

How can one become a professional computer vision engineer, especially considering the significant changes in the steps and strategies to enter the machine learning industry in a short period compared to the past?

Artificial Intelligence Advances at an Innovative Pace

You might not be surprised by the constantly changing nature of the ML industry, as you are well aware that artificial intelligence is advancing at a rapid pace. For most ML practitioners, the methods used to secure a position in 2020 may not necessarily work in 2021. We all need to adapt.

This article will introduce eight methods you can explore and utilize to start your journey to becoming a computer vision engineer.

1. Massive Open Online Courses (MOOCs)

Learning Path for Computer Vision Engineers

MOOC poster, created by Mathieu Plourde on April 4, 2013, licensed under CC-by on Flickr, discussing the meaning of “Massive Open Online Courses” or MOOCs.

MOOCs, launched in 2008, are a modern learning tool and are currently the preferred method for data scientists and machine learning practitioners to gain domain knowledge, often accompanied by recognized certifications and diplomas.

In most cases, MOOCs are much cheaper than traditional methods of learning through academic institutions and universities. It is not uncommon for MOOC courses to provide financial support for students through discount programs.

Another advantage of MOOCs over academic institutions is the flexibility to attend courses, online classes, and exams at times that suit you, rather than following a compact schedule structure and exam time allocation.

“Education is not about herding sheep. Education is about helping each student succeed.” — Andrew Ng

There are numerous MOOCs related to computer vision available online; the primary barrier you may face is choosing the right online course for you.

It is recommended to observe the skills and technical requirements of job seekers at the beginning before selecting suitable online courses.

Here are some links to MOOCs:

Become a Computer Vision Expert

https://www.udacity.com/course/computer-vision-nanodegree–nd891

Deep Learning Specialization

https://www.coursera.org/specializations/deep-learning

Advanced Computer Vision with TensorFlow

https://www.coursera.org/learn/advanced-computer-vision-with-tensorflow

Introduction to TensorFlow for Deep Learning

https://www.udacity.com/course/intro-to-tensorflow-for-deep-learning–ud187

DeepLearning AI TensorFlow Developer Professional Certificate

https://www.coursera.org/professional-certificates/tensorflow-in-practice

Introduction to Computer Vision

https://www.udacity.com/course/introduction-to-computer-vision–ud810

Basics of Computer Vision

https://www.coursera.org/learn/computer-vision-basics

Practical Deep Learning for Coders

https://course.fast.ai/

There are many MOOC courses available, but when searching, consider the following factors: job relevance, technical requirements, duration, course reviews, and cost.

MOOCs are not just for beginners; experienced ML practitioners and deep learning engineers also take intermediate and advanced MOOCs on specific computer vision-related topics to enhance their skills or gain domain expertise.

In 2021, speed, applicability, and practicality are the focus for ML practitioners. To transition from a machine learning student to a professional, choosing MOOCs that focus on the practical aspects of ML rather than the theoretical content taught in universities can save time and cost.

Note that pursuing an advanced degree at a university has its benefits; you should conduct your own in-depth research before making career decisions.

2. Machine Learning Libraries and Frameworks

Computer vision as a field encompasses many methods and techniques to solve common CV problems such as object detection, facial recognition, pose estimation, etc.

You typically do not need to reinvent the wheel or develop new algorithms to solve trivial computer vision tasks (unless you are working in the research field). Many tools you use in your learning and career can be obtained through ML libraries and frameworks.

ML libraries and frameworks provide a set of tools for ML practitioners to implement, train, test, and deploy computer vision solutions. Popular tools and libraries include TensorFlow and PyTorch, as well as FastAI, Caffe2, Keras, Scikit Learn, MXnet, Darknet, etc.

The ML industry seems to have established TensorFlow (Keras) and PyTorch as industry standards. As a computer vision engineer, I extensively use the TensorFlow platform to develop ML models for different environments.

Here is a typical checklist you can use as a guide when learning ML libraries:

Load structured datasets for training
Preprocess various forms of datasets (text, images, numbers)
Perform data augmentation
Implement a neural network from scratch
Implement a deep neural network from scratch
Implement a custom training process for the network
Perform image classification using various network architectures
Load models for transfer learning
Implement custom layers, dropout layers, and common layers in neural networks from scratch
Train deep learning models
Monitor the training process with TensorBoard
Save and load trained models.

Note: The above list does not encompass everything you need to know; in fact, I suspect it only touches on 20% of what you should cover to ensure you are prepared to take on the role of a professional CV engineer. Nevertheless, the above list is merely a guideline, and you are free to add and modify it as you see fit.

3. Reading Books

To become a computer vision engineer in 2021 and beyond, you must refer to practical machine learning and computer vision books as learning resources.

Computer vision engineers never stop learning, primarily because the field of artificial intelligence is advancing daily.

While maintaining their careers, it is common to encounter ML practitioners at the same learning level as students, who still need to read books, blogs, research papers, and articles to ensure they do not fall behind in the ML industry.

A highly recommended book for CV engineers is Aurélien Géron’s Machine Learning with Scikit Learn, Keras, and TensorFlow. This book is suitable for all ML practitioners, from data scientists to NLP engineers.

Machine Learning with Scikit Learn, Keras, and TensorFlow

It should be clear that reading practical books should be part of your ongoing personal learning strategy.

Suppose you find a specific area in your chosen field where you may lack expertise, whether it be mathematics, statistics, programming, or algorithms. In that case, there are many practical books that are easy to learn and effective for ML practitioners at all levels.

4. Cloud Services

Should computer vision engineers understand cloud services?

Yes, but you do not need to become an expert; there are cloud and data engineers whose sole responsibility is to specialize in cloud solutions and services.

Nevertheless, it is essential for CV engineers to understand how to run machine learning models on cloud services such as GCP, Microsoft Azure, and AWS.

Here are some reasons CV engineers should understand and utilize cloud services:

Availability of Computing Resources. Training deep learning models can be costly, especially if you have to purchase workstations and machines worth thousands of dollars. Cloud services offer GPUs with different computing specifications for running specific jobs. These computing resources are billed by the hour.
Remote Access to Shared Workspaces. Most teams utilize cloud services for online development environments to ensure every team member can access remote workspaces and resources.
Ready-made Solutions and Pre-trained Models.
Platforms that serve machine learning models through APIs, etc..

Using cloud services can be daunting, and sometimes costs can be unexpected—especially if you forget to shut down instances. Having knowledge of cloud services will indeed give you an edge among CV engineers in this industry.

5. Certifications

This refers not to the certificates or credentials obtained from participating in online courses. Certifications offered to ML practitioners can demonstrate expertise within specific libraries, cloud services, and frameworks.

TensorFlow, AWS, and GCP are tools and resources mentioned in this article. Another point is that companies and organizations providing these tools and services also issue recognized certificates.

A few years ago, having a portfolio with an advanced degree was sufficient to secure ML roles. With the influx of ML practitioners, it seems likely that we will see lists of candidates filtered by certifications among recruiters and employers.

Here is a list of certifications relevant to ML practitioners:

TensorFlow Developer Certificate

https://www.tensorflow.org/certificate

Google Cloud ML Engineer Certificate

https://cloud.google.com/certification/machine-learning-engineer

AWS Certified Machine Learning Specialist

https://aws.amazon.com/certification/certified-machine-learning-specialty/

Google Cloud Professional Data Engineer

https://cloud.google.com/certification/data-engineer

Google Cloud Associate Cloud Engineer

https://cloud.google.com/certification/cloud-engineer

DeepLearning.AI TensorFlow Developer Professional Certificate

https://www.coursera.org/professional-certificates/tensorflow-in-practice

From this point onwards, the information contained about becoming a computer vision engineer is evident. However, you will find some content very valuable.

6. Deep Learning

Deep learning is a field that uses deep artificial neural networks to detect data patterns.

Computer vision engineers typically utilize deep learning models to solve CV tasks. To put it mildly, CV engineers must understand the fundamental concepts and ideas of the DL field.

By 2021, deep learning will shift from primarily using convolutional neural networks as building blocks to the recently introduced transformer architecture.

ML practitioners are likely not required to understand how to implement and utilize transformers for computer vision tasks—

ML practitioners are likely not required to understand how to implement and utilize transformers for computer vision tasks, as this remains a research area.

Some proactive ML practitioners are now exploring transformers and learning how to implement and apply them.

7. Mobile and Edge Devices

Instagram, TikTok, YouTube, Facebook… these are companies with mobile applications where you will find ML models being utilized in some form.

In 2021, computer vision engineers must understand the tools and frameworks that enable the development and integration of models in mobile environments.

CV engineers should be familiar with the following platforms, tools, and frameworks:

TensorFlow Lite

https://www.tensorflow.org/lite

Core ML

https://developer.apple.com/documentation/coreml

Apple Vision Framework

https://developer.apple.com/documentation/vision

TensorFlow React

https://blog.tensorflow.org/2020/02/tensorflowjs-for-react-native-is-here.html

Create ML

https://developer.apple.com/machine-learning/create-ml/

A useful learning resource for understanding the application of computer vision and deep learning in mobile environments is “Practical Deep Learning for Cloud, Mobile, and Edge”: https://www.amazon.co.uk/Practical-Learning-Cloud-Mobile-Hands/dp/149203486X

8. Programming Languages

Any form of software engineering profession requires at least one programming language to be understood.

Typically, computer vision engineers are proficient in Python. You will likely find that Python is the most required programming language in most machine learning-related job roles.

By 2021, CV engineers will need to have proficiency in at least one language and intermediate proficiency in two or three other languages.

The reason is that deep learning models are applied across different platforms and environments, each of which utilizes other core programming languages to perform functions and operations.

I am currently integrating deep learning models in mobile environments, which requires me to be proficient in Python, JavaScript, and Swift. In 2021, I will expand my programming language skills to include Kotlin and Java.

The trick here is not to memorize all available syntax in one language and then move on to the next. Instead, it is more important to understand the basic principles of object-oriented programming and the coding patterns used in most programming languages.

Conclusion

The demand for computer vision/deep learning engineers is likely to increase with the adoption of more camera-oriented AI applications.

What ML practitioners learn to become CV engineers can be quite extensive, but once you start learning, complete some MOOCs, and learn core ML libraries, you will find the learning process becoming easier.

This article has introduced eight methods you can take and actions to become a CV engineer in 2021. In summary, here is a recap list:

Take online courses to gain expertise or enhance knowledge.
Utilize machine libraries and frameworks.
Read practical ML/DL books
Understand cloud services like GCP, AWS, etc.
Consider obtaining certifications for certain tools and libraries
Understand the basics of deep learning
Learn tools, libraries, and frameworks that can integrate deep learning models in mobile environments.
Understand programming patterns and principles, such as object-oriented programming.

Good news!
Beginner Learning Vision Knowledge Planet
Is now open to the public👇👇👇






Download 1: OpenCV-Contrib Extension Module Chinese Version Tutorial
Reply “Extension Module Chinese Tutorial” in the "Beginner Learning Vision" public account background to download the first OpenCV extension module tutorial in Chinese on the internet, covering installation of extension modules, SFM algorithms, stereo vision, object tracking, biological vision, super-resolution processing, and more than twenty chapters.

Download 2: Python Vision Practical Project 52 Lectures
Reply “Python Vision Practical Project” in the "Beginner Learning Vision" public account background to download 31 visual practical projects including image segmentation, mask detection, lane detection, vehicle counting, eyeliner addition, license plate recognition, character recognition, emotion detection, text content extraction, face recognition, etc., to help quickly learn computer vision.

Download 3: OpenCV Practical Projects 20 Lectures
Reply “OpenCV Practical Projects 20 Lectures” in the "Beginner Learning Vision" public account background to download 20 practical projects based on OpenCV for advanced learning of OpenCV.

Group Chat

Welcome to join the reader group of the public account to communicate with peers. Currently, there are WeChat groups for SLAM, 3D vision, sensors, autonomous driving, computational photography, detection, segmentation, recognition, medical imaging, GAN, algorithm competitions, etc. (will gradually be subdivided). Please scan the WeChat number below to join the group, remark: “Nickname + School/Company + Research Direction”, for example: “Zhang San + Shanghai Jiao Tong University + Visual SLAM”. Please follow the format for remarks; otherwise, it will not be approved. After successful addition, you will be invited to related WeChat groups based on research direction. Please do not send advertisements in the group; otherwise, you will be removed from the group. Thank you for your understanding~