Question about "Exercise 3 - yolo_non_max_suppression"

Hi friends and mentors,

I am in the “Exercise 3 - yolo_non_max_suppression” from hw Autonomous_driving_application_Car_detection. I am not sure I passed this section or not. But some questions here.

Q1. This is what I have, but do I pass or not?

Q2. The code below is correct or not? I found out that, I still get the same index (nms_indices) if I use the constant max_boxes=10 directly. Another thing is, I didn’t use the iou_threshold at all (because the hint says the tf version for this hw is not supported). I probably already made some mistakes already?
{code removed by mentor}
thank you!

Hello @sunson29,

Yes! It said “All tests passed!”, right? However, there will be other tests when you submit your assignment, so whether it can pass those tests are for you to tell us :wink:

I just read it but the hint is talking about “score_threshold” instead of “iou_threshold”. Would you like to try it again first?


PS: I am removing the code for you because we can’t share it here.

You don’t know if you pass until you submit it for grading.

Q1, I put iou_threshold in the function tf.image.non_max_suppression, and nothing happens (at least no error). From the reference, what is this score_threshold ? Is this same as threshold in yolo_filter_boxes(boxes, box_confidence, box_class_probs, threshold = .6) from Exercise 1 - yolo_filter_boxes ?

Q2, for the function tf.image.non_max_suppression , I tried directly using max_boxes = 10 instead of max_boxes_tensor = tf.Variable(max_boxes, dtype='int32') . both give me no error. So, I am not sure which is the correct way.

Thank you!

Hello @sunson29,

It is usual that we want to understand how to use a function provided by a package, and tensorflow has very good documentation, so I am suggesting you to try to read through it yourself. You can get yourself the answers of what you have asked - definition of score_threshold, and what is correct to set to max_output_size.

I encourage you to always check out the documentation for functions that you are not familiar with. This is a good practice.


1 Like