Grading of my Image Segmantation with Unet lab

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

Hello,
I just completed the Image Segmantation with Unet lab. It was very interesting and really helped me gain a better understanding of Unet.

I passed all three parts of the lab without any issues. However, when I submitted the lab I received a passing grade but with a score of only 66%.

Since I passed and can now move forward its not a big issue. Still, it would be greatly appreciated if someone there could please look at my code and just tell me what the problem is. I would like to receive a grade of 100%.

2

We cannot see your code directly. It’s also against the rules to publish it here on a public thread. I will send you a private DM message about how to proceed.

But it might help just to see the full grader output. Which section is it that gets 0 points and does the “Show grader output” link give any further info?

Also note as a general matter that just passing the test cases in the notebook does not prove your code is fully correct. If you then fail the grader, it probably means there is something about your code that is not fully general. Meaning that you’ve wired in some assumptions that just happen to match the test cases in the notebook, but fail when the grader uses a different test case.

3

Ok, I just looked at that notebook again to refresh my memory. Here’s one way to get the behavior you are seeing: hard code 32 everywhere in unet_model, instead of using the actual formal parameter n_filters. What if the grader passes a different value than the default for n_filters?

1 Like
4

Hi Paul - I just saw this email. I’ll try hard coding n_filters to 32 and see if that works. If it does, I’ll let you know.
Marc

1 Like
5

Just to make sure we are on the same page, my point is that hard-coding is a mistake, right? Sorry if my previous post was not as clear as I intended.

1 Like
6

I did misunderstand that.
The first download is then the correct version - or at least the best I can do.
Marc

7

Paul - I just resubmitted and all 3 parts are now correct.
The problem was that I DID hardcode the number of classes.
Thanks so much for your help.
Marc

2 Likes
8

That is great news that you found the solution. A valuable “gotcha” to remember. Onward! :nerd_face: