Your robot can now run software, move its joints, and keep track of its own frames. What it still cannot do is understand what is around it. Perception is how a robot turns raw sensor data into information it can act on, and for most robots the richest source of that data is a camera.
You'll start with the basics of working with images in OpenCV: reading frames, changing color spaces, finding edges and contours. Then you'll learn how a camera maps the 3D world onto a flat image, and how camera calibration measures the intrinsic parameters you need to trust those pixels. From there you'll see how two cameras can recover depth with stereo vision, so the robot knows how far away things are.
Finally you'll meet convolutional neural networks, the models behind modern object detection and recognition. These ideas connect directly back to ROS 2 image topics and TF2 frames, and they feed into Motion Planning & Navigation and Machine Learning & Simulation.
A beginner course on OpenCV in Python, from reading images and video to color spaces, edge detection, and contours.
A short explanation of what camera calibration is and how the intrinsic parameters of a camera are estimated.
How two images taken from different viewpoints let a robot recover depth and 3D information about a scene.
A quick introduction to CNNs, the building block behind modern image recognition and object detection.