최종 Bbox의 Class Score 계산

• 하나의 Bbox Class Score = Class Score x Object Confidence

  

  출처: https://www.slideshare.net/TaegyunJeon1/pr12-you-only-look-once-yolo-unified-realtime-object-detection

• 모든 Bbox에 대해 위 작업 진행

  

  출처: https://www.slideshare.net/TaegyunJeon1/pr12-you-only-look-once-yolo-unified-realtime-object-detection

Filtering → Sorting → NMS

• 전체 과정 (Dog Class에 대해서만)

  

  출처: https://www.slideshare.net/TaegyunJeon1/pr12-you-only-look-once-yolo-unified-realtime-object-detection

• NMS (Non-Maximal Surpression)

  

  출처: https://www.slideshare.net/TaegyunJeon1/pr12-you-only-look-once-yolo-unified-realtime-object-detection

Draw Bounding Box

출처: https://www.slideshare.net/TaegyunJeon1/pr12-you-only-look-once-yolo-unified-realtime-object-detection