aakash basnet - Image classification

Import the fashion MNIST dataset

Code

import tensorflow as tf
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

print(tf.__version__)

fashion_data = tf.keras.datasets.fashion_mnist
(train_data, train_label) , (test_data, test_label) = fashion_data.load_data()

2.15.0

Preprocessing the data

Code

train_data.shape

(60000, 28, 28)

Code

test_label.shape

(10000,)

Code

plt.figure()
plt.imshow(train_data[0])
plt.colorbar()
plt.grid(False)
plt.show()

Code

x_valid, x_train = train_data[:5000] / 255.0, train_data[5000:]/255.0
y_valid, y_train = train_label[:5000], train_label[5000:]

Code

class_map = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']

Code

plt.figure(figsize=(10,10))
for i in range(25):
    plt.subplot(5,5, i+1)
    plt.xticks([])
    plt.yticks([])
    plt.xlabel(class_map[train_label[i]])
    plt.imshow(train_data[i])
plt.show()

Building the model using neural network

Sequential Model is the simplest form of Model that keras library provides. It is composed of a single stack of layer connected sequentially. Also, known as sequential API. The first layer in the model is is Flatten model ,which simply flattens the 2D image array to 1D. Then, the Dense layer with 300 neurons which uses Relu as an activation function is stacked. Then, there is second Dense layer with 100 neurons , with same Relu as activation function. The final layer in the model stack is output layuer, which is also a Dense layer with 10 neurons (one per class) and it uses softmax function for multi-class classification.

Code

input_layer = tf.keras.layers.Flatten(input_shape=(28,28))
dense_layer1 = tf.keras.layers.Dense(300, activation='relu')
dense_layer2 = tf.keras.layers.Dense(100, activation='relu')
output_layer = tf.keras.layers.Dense(10, activation = 'softmax')

model = tf.keras.Sequential([input_layer,dense_layer1, dense_layer2 ,output_layer])
optimizer = 'adam'
loss_function = tf.keras.losses.SparseCategoricalCrossentropy()
metrics = ['accuracy']

model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 flatten (Flatten)           (None, 784)               0         
                                                                 
 dense (Dense)               (None, 300)               235500    
                                                                 
 dense_1 (Dense)             (None, 100)               30100     
                                                                 
 dense_2 (Dense)             (None, 10)                1010      
                                                                 
=================================================================
Total params: 266610 (1.02 MB)
Trainable params: 266610 (1.02 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________

Training the model

Code

model.compile(optimizer=optimizer, loss= loss_function, metrics=metrics )
history = model.fit(x_train,y_train, epochs=10, validation_data = (x_valid, y_valid))

Epoch 1/10
1719/1719 [==============================] - 2s 852us/step - loss: 0.4921 - accuracy: 0.8241 - val_loss: 0.3733 - val_accuracy: 0.8646
Epoch 2/10
1719/1719 [==============================] - 1s 788us/step - loss: 0.3670 - accuracy: 0.8647 - val_loss: 0.3491 - val_accuracy: 0.8706
Epoch 3/10
1719/1719 [==============================] - 2s 891us/step - loss: 0.3282 - accuracy: 0.8789 - val_loss: 0.3108 - val_accuracy: 0.8846
Epoch 4/10
1719/1719 [==============================] - 1s 775us/step - loss: 0.3058 - accuracy: 0.8853 - val_loss: 0.3005 - val_accuracy: 0.8934
Epoch 5/10
1719/1719 [==============================] - 1s 822us/step - loss: 0.2847 - accuracy: 0.8929 - val_loss: 0.3024 - val_accuracy: 0.8876
Epoch 6/10
1719/1719 [==============================] - 1s 769us/step - loss: 0.2725 - accuracy: 0.8971 - val_loss: 0.3025 - val_accuracy: 0.8932
Epoch 7/10
1719/1719 [==============================] - 1s 799us/step - loss: 0.2566 - accuracy: 0.9029 - val_loss: 0.2946 - val_accuracy: 0.8966
Epoch 8/10
1719/1719 [==============================] - 1s 776us/step - loss: 0.2473 - accuracy: 0.9064 - val_loss: 0.3052 - val_accuracy: 0.8944
Epoch 9/10
1719/1719 [==============================] - 1s 784us/step - loss: 0.2394 - accuracy: 0.9109 - val_loss: 0.3212 - val_accuracy: 0.8896
Epoch 10/10
1719/1719 [==============================] - 1s 818us/step - loss: 0.2280 - accuracy: 0.9131 - val_loss: 0.3100 - val_accuracy: 0.8936

Code


pd.DataFrame(history.history).plot()
plt.grid(True)
plt.gca().set_ylim(0,1)
plt.show()

Testing the model

Code

test_loss, test_acc = model.evaluate(test_data, test_label)

313/313 [==============================] - 0s 433us/step - loss: 55.4865 - accuracy: 0.8696