Attempting an evaluation.py file for a handwritten digit recognizer

Here's the code i have so far:

# Handwritten digit recognition for MNIST dataset using Convolutional Neural Networks

# Step $1$: Import all required keras libraries

from keras.models import load$\_$model # This is used to load your saved model

from keras.datasets import mnist # This is used to load mnist dataset later

from keras.utils import np$\_$utils # This will be used to convert your test image to a categorical class $($digit from $0$ to $9)$

# Step $2$: Load and return training and test datasets

def load$\_$dataset$()$:

# $2$a$.$ Load dataset X$\_$train, X$\_$test, y$\_$train, y$\_$test via imported keras library

$($X$\_$train, y$\_$train$),($X$\_$test, y$\_$test$)=$ mnist.load$\_$data$()$

X$\_$train $=$ X$\_$train.reshape$(($X$\_$train.shape$[0],28,28,1)).$astype$(\text{'}$float$32\text{'})$

# $2$b$.$ reshape for X train and test vars $-$ Hint: X$\_$train $=$ X$\_$train.reshape$(($X$\_$train.shape$[0],28,28,1)).$astype$(\text{'}$float$32\text{'})$

X$\_$test $=$ X$\_$test.reshape$(($X$\_$test.shape$[0],28,28,1)).$astype$(\text{'}$float$32\text{'})$

# $2$c$.$ normalize inputs from $0-255$ to $0-1-$ Hint: X$\_$train $=$ X$\_$train $/255$

X$\_$train $=$ X$\_$train $/255\mathrm{.}0$

X$\_$test $=$ X$\_$test $/255\mathrm{.}0$

# $2$d$.$ Convert y$\_$train and y$\_$test to categorical classes $-$ Hint: y$\_$train $=$ np$\_$utils.to$\_$categorical$($y$\_$train$)$

y$\_$train $=$ np$\_$utils.to$\_$categorical$($y$\_$train$)$

y$\_$test $=$ np$\_$utils.to$\_$categorical$($y$\_$test$)$

# $2$e$.$ return your X$\_$train, X$\_$test, y$\_$train, y$\_$test

return X$\_$train, X$\_$test, y$\_$train, y$\_$test

# Step $3$: Load your saved model

from tensorflow.keras.models import load$\_$model

model $=$ load$\_$model$(\text{'}$digitRecognizer$.$h$5\text{'})$

# Step $4$: Evaluate your model via your$\_$model$\_$name.evaluate$($X$\_$test, y$\_$test, verbose $=0)$ function

test$\_$loss, test$\_$acc $=$ model.evaluate$($test$\_$images, test$\_$labels$)$

print$($f$\text{'}$Test accuracy: $\{$test$\_$acc$\}\text{'})$

# Code below to make a prediction for a new image.

predictions $=$ model.predict$($test$\_$images$)$

# Step $5$: This section below is optional and can be copied from your digitRecognizer.py file from Step $8$ onwards $-$ load required keras libraries

from keras.preprocessing.image import load$\_$img

from keras.preprocessing.image import img$\_$to$\_$array

# Step $6$: load and normalize new image

def load$\_$new$\_$image$($path$)$:

# $6$a$.$ load new image

newImage $=$ load$\_$img$($path$,$ grayscale$=$True, target$\_$size$=(28,28))$

# $6$b$.$ Convert image to array

newImage $=$ img$\_$to$\_$array$($newImage$)$

# $6$c$.$ reshape into a single sample with $1$ channel $($similar to how you reshaped in load$\_$dataset function$)$

# $6$d$.$ normalize image data $-$ Hint: newImage $=$ newImage $/255$

# $6$e$.$ return newImage

# Step $7$: load a new image and predict its class

def test$\_$model$\_$performance$()$:

# $7$a$.$ Call the above load image function

img $=$ load$\_$new$\_$image$(\text{'}$your$\_$new$\_$image$\_$file$\_$path'$)$

# $7$b$.$ load your CNN model $($digitRecognizer$.$h$5$ file$)$

# $7$c$.$ predict the class $-$ Hint: imageClass $=$ your$\_$model$\_$name.predict$\_$classes$($img$)$

# $7$d$.$ Print prediction result

print$($imageClass$[0])$

# Step $8$: Test model performance here by calling the above test$\_$model$\_$performance function