Closed ronanbyrne96 closed 4 years ago
Dear ronanbyrne96, this is a good point. In this example the training of the NN has been done only with images taken with numbers that are in front position. Thus in the training dataset we do not have numbers that are half visible (i.e. images of a half 4 because the counter is turning to a 5), this is why we skip the least significant number while checking results with the GT.
I'm adding a comment to explain this in the code.
Thanks
Small fix needed in applications/WaterMeter/Mnist.c:350
should be
So it checks that the last digit is correct.