In 

[ ]:  import numpy as np

import tensorflow
import keras
import PIL

In [ ]:  import tensorflow as tf

import matplotlib.pyplot as plt

In [ ]:  from tensorflow import keras

from tensorflow.keras import layers,datasets, models
from tensorflow.keras.models import Sequential

In [ ]:  (train_images, train_labels),(test_images, test_labels)=datasets.mnist.load_

​
train_images = train_images.reshape((60000, 28, 28, 1))
test_images =test_images.reshape((10000, 28, 28,1))
​
train_images, test_images= train_images/ 255.0, test_images/ 255.0
​
print("TRAIN IMAGES: ", train_images.shape)
print("TEST IMAGES: ", test_images.shape)

Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-da

tasets/mnist.npz (https://storage.googleapis.com/tensorflow/tf-keras-datase
ts/mnist.npz)

11493376/11490434 [==============================] - 0s 0us/step

11501568/11490434 [==============================] - 0s 0us/step

TRAIN IMAGES: (60000, 28, 28, 1)

TEST IMAGES: (10000, 28, 28, 1)

In [ ]:  num_classes = 10
img_height = 28
img_width = 28
​
model = Sequential([
layers.Conv2D(64, (3, 3), activation= 'relu', input_shape=(28,28,1)),
layers.Conv2D(32, 3, padding='same', activation= 'relu'),
layers.MaxPooling2D(),
layers.Conv2D(16, 3, padding='same', activation='relu'),
layers.MaxPooling2D(),
layers.Conv2D(64, 3, padding='same', activation='relu'),
layers.MaxPooling2D(),
layers. Flatten(),
layers.Dense (128, activation='relu'),
layers.Dense (10, activation='sigmoid')
])

In [ ]:  model.compile(optimizer='adam',
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=
metrics=['accuracy'])
In [ ]:  model.summary()

Model: "sequential"

_________________________________________________________________

Layer (type) Output Shape Param #

=================================================================

conv2d (Conv2D) (None, 26, 26, 64) 640

conv2d_1 (Conv2D) (None, 26, 26, 32) 18464

max_pooling2d (MaxPooling2D (None, 13, 13, 32) 0

conv2d_2 (Conv2D) (None, 13, 13, 16) 4624

max_pooling2d_1 (MaxPooling (None, 6, 6, 16) 0

2D)

conv2d_3 (Conv2D) (None, 6, 6, 64) 9280

max_pooling2d_2 (MaxPooling (None, 3, 3, 64) 0

2D)

flatten (Flatten) (None, 576) 0

dense (Dense) (None, 128) 73856

dense_1 (Dense) (None, 10) 1290

=================================================================

Total params: 108,154

Trainable params: 108,154

Non-trainable params: 0

_________________________________________________________________

In [ ]:  epochs=10
history=model.fit(
train_images,
train_labels,
epochs=epochs
)

Epoch 1/10

/usr/local/lib/python3.7/dist-packages/tensorflow/python/util/dispatch.py:1
082: UserWarning: "`sparse_categorical_crossentropy` received `from_logits=
True`, but the `output` argument was produced by a sigmoid or softmax activ
ation and thus does not represent logits. Was this intended?"

return dispatch_target(*args, **kwargs)

1875/1875 [==============================] - 18s 4ms/step - loss: 0.1353 -

accuracy: 0.9569

Epoch 2/10

1875/1875 [==============================] - 7s 4ms/step - loss: 0.0439 - a

ccuracy: 0.9862

Epoch 3/10

1875/1875 [==============================] - 7s 4ms/step - loss: 0.0309 - a

ccuracy: 0.9904

Epoch 4/10

1875/1875 [==============================] - 7s 4ms/step - loss: 0.0254 - a

ccuracy: 0.9918

Epoch 5/10

1875/1875 [==============================] - 13s 7ms/step - loss: 0.0214 -

accuracy: 0.9934

Epoch 6/10

1875/1875 [==============================] - 9s 5ms/step - loss: 0.0169 - a

ccuracy: 0.9945

Epoch 7/10

1875/1875 [==============================] - 9s 5ms/step - loss: 0.0162 - a

ccuracy: 0.9945

Epoch 8/10

1875/1875 [==============================] - 9s 5ms/step - loss: 0.0129 - a

ccuracy: 0.9960

Epoch 9/10

1875/1875 [==============================] - 10s 5ms/step - loss: 0.0130 -

accuracy: 0.9959

Epoch 10/10

1875/1875 [==============================] - 7s 4ms/step - loss: 0.0109 - a

ccuracy: 0.9963

In [ ]:  acc=history.history['accuracy']
loss=history.history['loss']
epochs_range=range(epochs)
​
plt.figure(figsize=(8,8))
plt.plot(epochs_range, acc, label='Training Accuracy')
plt.plot(epochs_range, loss, label='Loss')
plt.legend(loc='lower right')
plt.title('Training Accuracy and Loss')

Out[10]: Text(0.5, 1.0, 'Training Accuracy and Loss')

In [ ]:  image=(train_images[1]).reshape(1,28,28,1)
​
model_pred=model.predict(image)
classes_x=np.argmax(model_pred, axis=1)
​
plt.imshow(image.reshape(28,28))
print(classes_x)

[0]

In [ ]:  image=(train_images[2]).reshape(1,28,28,1)
​
model_pred=model.predict(image)
classes_x=np.argmax(model_pred, axis=1)
​
plt.imshow(image.reshape(28,28))
print(classes_x)
​

[4]

In [ ]:  model.save("tf-cnn-model.hs")

In [ ]:  loaded_model=models.load_model("tf-cnn-model.hs")

In [ ]:  image=(train_images[2]).reshape(1,28,28,1)
​
model_pred=model.predict(image)
classes_x=np.argmax(model_pred, axis=1)
​
plt.imshow(image.reshape(28,28))
print(classes_x)

[4]

In [ ]:  ​