Computer vision (also called machine vision) is one of the most exciting frontiers in computing. With computer vision, anyone with the right skills can solve problems that used to take years of research to solve. You can build a model to detect defects in products, to identify wildfires, or to detect the wildlife in a given area. If a solution benefits from identifying or tracking objects in video or finding objects in an image, computer vision can be helpful.
Here, we have curated our best free learning resources into a Roboflow Computer Vision Course to help you start learning computer vision.
Before you begin building your first model, you first need to learn the fundamentals of computer vision. In this section, we list tutorials and guides that discuss what computer vision is, what classes of problems can be solved with computer vision, and what the process looks like to build a successful computer vision model. The knowledge you acquire in this section will set you in good stead to solve your first model with computer vision.
Learn about what computer vision is and what problems you can solve with computer vision.
All the acronyms! In this video, you'll learn about Artificial Intelligence, Machine Learning, Computer Vision, and how they compare.
Learn about the spectrum of classification, object detection, and segmentation computer vision models and when they are useful. (Don't worry, we define all these terms in the video!)
Learn about the process to define a problem statement and build a model based on your requirements.
Train your first computer vision model along with us in our Roboflow training guide.
Build the knowledge you need to properly define a computer vision problem and set yourself up for success in your building.
Learn what image classification is, what problems it solves, and when you might want to use image classification.
In this section, we have aggregated tutorials that will be useful as you begin training your first model. We answer common questions like "What is Mean Average Precision?" and provide key knowledge about how to use preprocessing and augmentation to build a more accurate model. We also have resources on evaluating models so you'll know the metrics that reflect the performance of your model.
Learn how many images you need to train a computer vision model.
Learn what Mean Average Precision is and how it is used to evaluate computer vision models.
Understand what precision and recall are and what these metrics tell you about model performance.
Learn about key preprocessing and augmentation techniques you can use to improve model performance.
Learn how to use Roboflow to assess the quality of your dataset and improve the model trained on your data.
Once you have built a model, the next task is to deploy it into production for use in the real world. In this section, we have aggregated tutorials for deploying computer vision models, providing you with the resources you need to start using your models. You can deploy models across many devices, from "edge deployment" cameras like the Luxonis OAK all the way to your browser.
How do you deploy a computer vision model? We answer that question in this video.
Learn how to deploy a custom computer vision model to a Luxonis OAK camera.
Learn how to deploy the model you have been working on to a Raspberry Pi.
Computer vision has use cases across dozens of industries. For example, construction sites can use computer vision to prevent accidents by detecting when people are too close to moving vehicles. Manufacturers of parts for cars can use computer vision to identify defects in products when they are on the assembly line. But, you may be wondering: what small and practical projects can I build to test my skills?
Below, we have curated some of our favorite tutorials on projects you can build with computer vision at home. However, do not let the ideas in the section below limit you. If you have a project idea that could benefit from being able to identify objects in an image or a video, you will be able to use your computer vision skills! There's no better way to reinforce what you have learned about computer vision than to build your own project that solves a problem you have.
There are many courses available online in addition to Roboflow's learning resources that you can follow to advance your knowledge of computer vision and related concepts. Below, we have aggregated some of our favorite courses that cover computer vision and machine learning concepts that will apply to your computer vision work.
Made with ML is an online course that covers how to "responsibly develop, deploy, and maintain ML." This course is a great way to learn about the fundamentals principles used across machine learning and computer vision, from regressions to data augmentation.
A free course designed for people with some coding experience, who want to learn how to apply deep learning and machine learning to practical problems.
In this course, you will learn about how to use the Luxonis OpenCV AI Kit and Roboflow to train and deploy a computer vision model.
There are many YouTube channels that publish videos on computer vision. Below, you can find a few of the many channels out there from which you can learn computer vision and stay up to date with the latest technologies in the industry.
Augmented Startups publishes regular tutorials on artificial intelligence and computer vision. On this channel, you'll learn about everything from object detection and tracking to using generative image models.
A channel run by OpenCV with dozens of videos on computer vision, ranging from using the OpenCV Python library to overviews on various computer vision models.
With software like Roboflow, you can build a computer vision model without any prior computer vision experience. Previously, learning computer vision involved an extensive investment of time and computing resources. Over the last few years, there have been advances in the field to make the technology more approachable. Now, you can use tools like Roboflow to build models hands-on without minimal to no code, which makes the learning process easier.
Learning hands-on computer vision with code is more difficult, however, often involving months of learning. This is an appropriate to take in your learning if you want to understand the "how" behind modern computer vision algorithms and build model architectures, infrastructure, or configurations from scratch.
You can build a computer vision algorithm in about a day with a tool like Roboflow, which handles the technical back-end and empowers you to focus more on solving a particular problem with computer vision. It takes a few days to learn about the different types of problems you can solve with computer vision and a few weeks to learn more fundamentals like improving model performance and deployment.
If you want to become a computer vision engineer, expect to spend a few months learning computer vision algorithm fundamentals, assuming you already have some prior knowledge of software engineering and mathematics. Your learning will involve learning about the structure of modern architectures, the evolution of computer vision, and the "how" and "why" behind today's state-of-the-art systems. Practical experience building models, optionally in a structured course or degree, will go a long way to help you building knowledge, too. You can expect to spend about a year building a solid foundation of the skills you'd need to start working on computer vision models for a business.