Getting started on Computer Vision

Computer vision is a rising field in Artificial Intelligence and it has become very powerful in terms of intelligently interpreting scenes and tracking objects. This blog article will help you get started on Computer Vision. I will list a few useful resources for self-learning, and also teach you how to set up the popular tools in computer vision.

First, it is a good idea to take some classes online for free at an introductory level to get yourself familiar with all the topics in computer vision and to get a good idea of what it can do:

Udacity Computer Vision class:

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

Cousera Robot Perception class:

https://www.coursera.org/learn/robotics-perception

Gradually, you will realize that machine learning is unavoidable in Computer Vision, so take the famous Andrew Ng Machine Learning class:

https://www.coursera.org/learn/machine-learning

Second, two excellent Computer Vision textbooks:

Computer Vision: Algorithms and Applications: https://www.amazon.com/Computer-Vision-Algorithms-Applications-Science/dp/1848829345/ref=sr_1_1?ie=UTF8&qid=1473621997&sr=8-1&keywords=computer+vision+algorithms+and+applications

Multiple View Geometry in Computer Vision: https://www.amazon.com/Multiple-View-Geometry-Computer-Vision/dp/0521540518/ref=sr_1_1?ie=UTF8&qid=1473622036&sr=8-1&keywords=multiview+geometry

The first one cited a lot of research papers. It is more like a guide to all the topics in computer vision, but you will not learn the actual content by just reading this book. It will guide you to other reading materials and papers that you need to find by yourself. The other one is very heavy on math. It will probably take a lot longer to finish, and it is totally OK if you don't finish it. It is much more fun to do some projects along the way and learn through exercise.

In short, computer vision takes in image (or a video), and it analyzes the image. With the right algorithm, it can tracking walking people in a surveillance video, recognize human faces or any object, read text, constructing a 3D model from pictures, and much more. Below are some examples of computer vision in action:

To be able to do all this cool stuff, you need software for programming. The usual tool for prototype code development is MATLAB. MATLAB does best in matrix operation, which is perfect because an image is a matrix to a computer. It also has a computer vision toolbox. If you are a university student, mostly likely it is free through the university account; otherwise they have a student discount for the software, and there is also a free alternative called Octave. For real time computing, OpenCV is a better choice, because it is optimized for computer vision.

Installing OpenCV is a little bit tricky, especially if you are new to computer science, because it is a third party library that needs to be manually linked to your program file every time you create a new project. It is easier on mac and harder on Visual Studio, but there are some YouTube videos that can help you out. I used to spend hours on setting up the environment on Visual Studio, but I am sure you can do better :)