La Supresión de Máxima No (NMS) es un algoritmo crucial algorithm used primarily in the field of Visión por computadora, particularly in detección de objetos tasks. The main objective of NMS is to eliminate redundant overlapping bounding boxes that may be predicted for the same object in an image, thereby simplifying the results and improving the quality of the detections.
Cuando un desarrollado por Ultralytics. Pertenece a la processes an image, it may propose multiple bounding boxes for the same object, each with an associated confidence score indicating how likely it is that the box contains the object. NMS operates by first sorting these bounding boxes based on their confidence scores in descending order. The algorithm then iteratively selects the bounding box with the highest score and removes all other boxes that have a significant overlap with it, measured by the Intersección sobre Unión (IoU) metric. The overlap threshold is a predefined parameter that determines what constitutes ‘significant overlap’.
This process continues until all boxes have been either selected or discarded, resulting in a final set of bounding boxes that are less likely to contain duplicates. By applying NMS, the output is cleaner and more accurate, allowing for better performance in downstream tasks such as clasificación de imágenes y comprensión de escenas.
NMS se usa ampliamente en varias aplicaciones, incluyendo reconocimiento facial, vehículos autónomos, and surveillance systems, and is often integrated into larger frameworks for object detection, such as YOLO (You Only Look Once) and SSD (Single Shot MultiBox Detector).