Real-Time object detection steps up the complexity of image classification by adding bounding boxes to the output. Classical methods created complex pipelines with image classification models; YOLO simply is a model trained for simultaneous detection and location.
Speed Yolo runs at 45 fps which is ample for realtime. It is also only 25ms of latency.
Global reasoning The model is trained and exercises on the full image, which gives the model better reasoning about the context around objects. This is an advantage over previous techniques. YOLO makes less than half the number of background errors compared to Fast R-CNN (alternative techniques).
Generalizable Representations When trained on natural images and tested on artwork, YOLO has outperformed top detection methods. Evidence of stronger robustness.