See our paper for more details on the full system. This makes it extremely fast, more than 1000x faster than R-CNN and 100x faster than Fast R-CNN.

It also makes predictions with a single network evaluation unlike systems like R-CNN which require thousands for a single image. It looks at the whole image at test time so its predictions are informed by global context in the image. Our model has several advantages over classifier-based systems. These bounding boxes are weighted by the predicted probabilities. This network divides the image into regions and predicts bounding boxes and probabilities for each region. We apply a single neural network to the full image. High scoring regions of the image are considered detections.

They apply the model to an image at multiple locations and scales. Prior detection systems repurpose classifiers or localizers to perform detection. Moreover, you can easily tradeoff between speed and accuracy simply by changing the size of the model, no retraining required!

5 IOU YOLOv3 is on par with Focal Loss but about 4x faster. On a Pascal Titan X it processes images at 30 FPS and has a mAP of 57.9% on COCO test-dev. You only look once (YOLO) is a state-of-the-art, real-time object detection system.

Download Coco