# You Only Look Once: Unified, Real-Time Object Detection

**Authors:** Joseph Redmon, Santosh Divvala, Ross Girshick, Ali Farhadi  
**arXiv:** [1506.02640](https://arxiv.org/abs/1506.02640)  
**Date:** 2015-06-08  
**Summary page:** https://www.summarizepaper.com/en/paper/1506.02640v1/

## Key points

- YOLO is a unified pipeline for object detection
- YOLO frames object detection as a regression problem
- YOLO predicts bounding boxes and class probabilities directly from full images using a single neural network in one evaluation
- YOLO can process images in real-time at a rate of 45 frames per second, making it much faster than existing detection systems
- YOLO utilizes global image context to detect and localize objects with reduced likelihood of background errors compared to top detection systems like R-CNN
- YOLO maintains moderate accuracy when detecting objects independently but can enhance performance by 2-3% points mAP when combined with state-of-the-art detectors
- The YOLO framework offers a highly efficient and accurate solution for real-time object detection, with its unified architecture and ability to optimize end-to-end directly on detection performance.

## Summary

The paper titled "You Only Look Once: Unified, Real-Time Object Detection" presents YOLO, a unified pipeline for object detection. Unlike previous approaches that repurpose classifiers for detection, YOLO frames object detection as a regression problem. It predicts bounding boxes and class probabilities directly from full images using a single neural network in one evaluation. One of the key advantages of YOLO is its speed; it can process images in real-time at an impressive rate of 45 frames per second which makes it hundreds to thousands of times faster than existing detection systems. Additionally, YOLO utilizes global image context to detect and localize objects with reduced likelihood of background errors compared to top detection systems like R-CNN. While YOLO achieves remarkable speeds, it also maintains moderate accuracy when detecting objects independently. However, when combined with state-of-the-art detectors, YOLO further enhances performance by 2-3% points mAP (mean average precision). Overall, the YOLO framework offers a highly efficient and accurate solution for real-time object detection. Its unified architecture and ability to optimize end-to-end directly on detection performance make it a significant advancement in the field.