Computer Vision, from Pixels to 3-D

About this course#

This course grew out of two university courses I took and the projects I built around them: Robotic Vision in the mechatronics graduate program at UFRN (Brazil) and Image Processing (TIM) at INSA Rouen Normandie (France), plus the computer vision I use in my research on autonomous drones and boats. I rewrote all of it as one path, in English, cleaned up the code, fixed the mistakes I made the first time, and added the explanations I wish I had had.

Each lesson follows the same rhythm:

1. The idea — what problem we are solving and the maths behind it, kept as short as possible.
2. From scratch — an implementation in plain Python/NumPy that you can read line by line.
3. The OpenCV way — the one-liner you would use in a real project, and a check that both agree.
4. Run it — every lesson is a single script. Run it, change a parameter, run it again.
5. Pitfalls — the bugs I actually hit, so you don't have to.

Who it is for#

Students and engineers who know some Python and want to understand computer vision instead of copying function calls. You need high-school maths; a bit of linear algebra (matrices, vectors) helps from Part I on and becomes essential in Part V. No deep learning is required: this is the classical toolbox that every vision pipeline — including the neural ones — still relies on for calibration, geometry, pre- and post-processing. The learning side — from k-NN and SVMs to CNNs, YOLO and diffusion models — is in the companion Machine Learning course.

How to follow it#

Start with lesson 00 to install the environment and download the code and data (one zip, about 20 MB). Parts I to IV build on each other; Part V (geometry and 3-D) needs lessons 11–14; the two case studies in Part VI reuse a bit of everything.