The CNN model does not reach 93% accuracy at all

[minqz2009]

The CNN model does not reach 93% accuracy at all. This model could only achieve an accuracy around 50~60%. The reason the accuracy from the output of the notebook in this repo is because, in the test function, they forgot to reinitialize the count and total.

So that 80% of test accuracy is dominant by the train accuracy.
Below is the real accuracy of the CNN model provided in this repo:

[goncinious]

Thanks for pointing this out @minqz2009! I'll correct it and update the repo as soon as I have time to improve it.

[minqz2009]

No worries mate