I wanted to know how the data from MNIST is preprocessed for the graded lab. I was trying to extend the lab by predicting the value from my own handwritten note and I can’t really figure out how the normalization might’ve happened since typically values are divided by 255 to create values between 0 and 1 but it doesn’t seem to be the case for the data I see.
Hello @arian81,
I didn’t prepare for that data so I do not know how it was preprocessed, however, if you wanted to have a model that can predict your own handwritten digits, you might download the MNIST from other sources such as this one which gets you with pixel values ranged between 0 and 255, normalize them, then train a new model with them.
Cheers,
Raymond
PS: Please try this after passing the assignment.
That’s what I did just preprocess mnist myself. Thanks for the help anyway