C4W3A1 - yolo_non_max_suppression

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

the example code for yolo_non_max_suppression use tf.image.non_max_suppression(boxes, scores, max_boxes, iou_threshold) to suppress the high IoU box.

While my understanding in this step, only the box with same class and overlap each other can be suppressed, that implies “classes” should be one parameter for this function of suppression, but there isn’t in the example code.

can anyone explain why is not used here for the suppression?

Thank you!

2

Hello there.
Please move your topic to the correct subcategory.
Here’s the community user guide to get started.

3

Suppose there was one vehicle object in the image. One predicted bounding box had an associated class prediction of car. Another predicted bounding box that almost entirely overlaps had a class prediction of truck. Which object(s) would you want to survive NMS and why?