Samuel I want to understand what Non-Maximum Suppression (NMS) means in computer vision. How does it help remove duplicate bounding boxes in object detection models? Can someone also explain how NMS improves detection accuracy?