Object detection locates objects with bounding boxes and classifies them.
Two-stage detectors: Region proposals, then classification. Faster R-CNN is the classic example. Accurate but slow.
One-stage detectors: Predict boxes and classes directly. YOLO, SSD. Faster, slightly less accurate.
YOLO: Divide image into grid. Each cell predicts boxes and class probabilities. Real-time detection.
Anchor boxes: Predefined box shapes. Network predicts offsets from anchors.
Interview question: "YOLO vs Faster R-CNN tradeoffs?"
YOLO is faster (real-time). Faster R-CNN is more accurate for small objects.