Level 3 · Detection, Segmentation & Motion

Classification answers "what is in this image"; this level answers the harder, more common real questions: where is it, what's its exact outline, how is it moving, and what text does it contain. You'll run and evaluate real pretrained object detectors and segmentation models, measure real optical flow between video frames, track a real object across a sequence, and run real OCR against real text in images.

By the end of this level you'll have built a real, multi-stage vision pipeline combining detection, tracking, and OCR against real images and video — the shape most production vision systems actually take.