The best way to make the modules stick is to build something. These five projects take you from OpenCV image processing through image classification with transfer learning, YOLO object detection, SAM 2 segmentation, and finally real-time object tracking on video.
Each one is a code-along video, so you can pause, type it out yourself and adapt it to your own images and datasets. Pick the project that matches the module you've just finished and come back for the rest.
Work through the OpenCV toolkit in Python: reading images and video, transformations, contours, color spaces and edge detection, finishing with a deep learning classifier.
Take a pretrained ResNet50, replace its classifier head and fine-tune it on your own images to build a custom image classifier.
Prepare a labelled dataset, then fine-tune YOLO11 for object detection and instance segmentation and run it on new images.
Fine-tune Segment Anything 2.1 on a custom dataset to produce precise pixel-level masks for the objects you care about.
Detect vehicles, follow them across video frames with ByteTrack, and estimate each vehicle's speed from how far it travels.