YOLO stands for "You Only Look Once." It's one of the most famous object detectors ever — fast enough to run on live video. The secret? Instead of scanning a picture a thousand times, it splits it into a grid and spots everything in a single glance. Let's see how.
YOLO lays a grid over the whole image. Each little cell has one job: "Is the center of an object in me? If so, what is it and how big?" Every cell answers at the same time — that's why it's so fast. The glowing cells are the ones "responsible" for an object.
Raw YOLO actually predicts tons of overlapping boxes — many for the same object, each with a confidence. Two clean-up steps fix it: (1) drop low-confidence boxes with a threshold, then (2) NMS — for a pile of overlapping boxes, keep the most confident one and delete its duplicates.
confidence threshold (drop boxes below this)
Watch the difference. The old way slides a window across every spot. YOLO looks once. The counter shows how many "looks" each one takes.
Our grid demo is the core idea. Here's the real YOLO world:
✅ Shows why 'one look' is fast. ⚠️ Simplified.
Ultralytics YOLOv8/v11 - runs in a few lines and detects 80+ objects in real time.
SSD, EfficientDet (fast), DETR (transformer, anchor-free). Pick by speed vs accuracy.
You now understand the world's most famous detector: YOLO splits the image into a grid, predicts everything in one look, then cleans up with a confidence threshold and NMS.