-
Notifications
You must be signed in to change notification settings - Fork 6.8k
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you for the fix! LGTM: )
@NRauschmayr Thank you for fixing the example. Can you please fix failing CI? |
I think we have to update the unit-test: it is using an outdated and not maintained model https://github.com/apache/incubator-mxnet/blob/a9458caec285d3c28d16588fef6aef5e219c7472/cpp-package/example/inference/unit_test_inception_inference.sh
|
@NRauschmayr Could you please resolve the merge conflicts?? |
@NRauschmayr Can you rebase with the latest master please ? @leleamol Gentle ping for review. |
Sorry for the late reply. I have not done the rebase yet, because I saw that the files I modified were deleted by the following commit #15164 I wanted to check first whether or not it makes sense to re-add those files, but I did not have the time yet to compile the latest c++ tutorial of the master branch. |
@mxnet-bot run ci [all] |
OK looks like this part is already refactored into the imagenet_inference.cpp example and it's handling the mean values correct. Still, thanks for proposing the patch. |
Description
Tutorial did not properly preprocess images and as such produce incorrect predictions. The example would set default values for mean but not for std_dev. So if the user does not provide the file that contains those values, the example would crash. The C++ example predicts now the same class as in the Python example. However, the probabilities differ (9.72 versus 10.411) and we should to look into that