YOLOv4 should be in the object detection results

According to the paper, YOLOv4 models are on the efficient frontier of the FPS/AP tradeoff, but they don’t appear anywhere on SOTA bench site.