Classification says what is in an image. Object detection goes further: it finds every object of interest, draws a box around it, and labels it. It powers self-driving perception, retail analytics, security cameras, and much else.
Detectors fall into a few families. Two-stage detectors such as Faster R-CNN first propose regions and then classify them. One-stage detectors such as YOLO predict boxes and classes in a single pass, which makes them fast enough for video. DETR treats detection as a set-prediction problem and uses a transformer. Along the way you will meet the core ideas they share: intersection over union and non-maximum suppression.
Finally, you will see how to train a detector on your own labelled images. Watch the lessons in order, then continue on to Segmentation.
A walkthrough of the two-stage detector family, showing how each version locates and identifies objects in an image.
An introduction to YOLO, the one-stage detector that predicts boxes and classes in a single pass.
How DETR uses a transformer to detect objects end to end, without hand-designed components such as anchors.
Training a YOLO detector on your own labelled images using the Ultralytics library.