YOLO
A family of fast object detectors that find all objects in an image with a single pass of a neural network, fast enough for live video.
Also You Only Look Once
Redmon, Divvala, Girshick and Farhadi, 2015
Older detectors looked at an image many times, once for each possible spot. YOLO looks once at the whole picture and calls out every object and its box in one go.
Where it fits
+All relationships as a list
Is a kind of / part of
Uses / built on
How it works
Before YOLO, the best detectors worked in two stages: propose thousands of candidate regions, then classify each one. Accurate, but slow.
YOLO (2015) turned detection into a single prediction problem. The image is divided into a grid; for each grid cell, one pass of a CNN predicts a few boxes, how confident it is that each box contains an object, and which class it is. Non-max suppression then removes duplicates.
The original paper reported 45 frames per second, and a smaller “Fast YOLO” at 155 frames per second. The name has since become a family: many versions by different teams have improved accuracy and speed, which is why “YOLO” today usually means “a fast one-stage detector in this lineage” rather than one specific model.
Everyday use cases
- Case 01
How you've probably met it
Many real-time camera applications (traffic monitoring, sports analytics, robotics demos) use a YOLO-family detector.
Strengths, weaknesses and when to use it
Strengths
- Real-time speed with good accuracy.
- Sees the whole image, so fewer false detections on background.
Weaknesses
- The original version struggled with small objects in groups and with precise localization.
When to use itWhen detection must run live, on video or edge devices.
Go deeper
Sources
Facts checked as of 2026-10-12