Pokemon Classification using Keras in Python

In this blog, We will learn about how to do Image classification using Keras in Python by classifying Pokemon. We will make a deep learning model that can classify different types of Pokemon. The dataset can be taken from this GitHub repository

This dataset contains 10 different pokemon such as:

Aerodactyl
Bulbasaur
Charmander
Dratini
Fearow
Meowth
Pikachu
Psyduck
Spearow
Squirtle

So, Let’s get started!!

Code

Firstly, we will import all the necessary python libraries in order to make our model.

#importing libraries
import os
import numpy as np
import  matplotlib.pyplot as plt
from keras.preprocessing import image
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import *
from keras.optimizers import adam

Data Preprocessing

We will make ImageDataGeneratot for processing the dataset. In, this we would rescale the data and apply data augmentation. You can read more about Data Augmentation from here. Here, we will apply various transformations in order to expand the training set as we have few images for training. We will change the size of the image to (64,64) so that there are fewer training parameters. Similarly, we will also change the shape of the test data, otherwise, our model would give an error.

from keras.preprocessing.image import ImageDataGenerator

#doing data augmentation
train_ = ImageDataGenerator(rescale=1/255.0,horizontal_flip=True,
                            rotation_range=0.2,shear_range=0.2,
                            height_shift_range=0.2,
                            width_shift_range=0.2,zoom_range=0.2)

train= train_.flow_from_directory(
    '../PROJECTS/pokemon/Train',
    target_size=(64,64),
    class_mode='categorical',
    
)

test_= ImageDataGenerator(rescale=1/255.0)
test=test_.flow_from_directory(
    '../PROJECTS/pokemon/Test (2)',
    target_size=(64,64),
    batch_size=16
)

Output: Found 1642 images belonging to 10 classes.
        Found 187 images belonging to 1 classes.

Now, we will build a deep learning model.

We will use the Sequential model with layers as Convolutional, Maxpooling, Flatten, and Dropouts to reduce overfitting.

The shape of the image passed is (64,64,3) as we have changed the size of the image to 64,64 and 3 is the number of channels as it is the colored image and has 3 channels RGB.

We will use Relu activations for the hidden layers and Softmax for the output layer.

np.random.seed(1000)
#Instantiate an empty model
model = Sequential()

# 1st Convolutional Layer
model.add(Conv2D(filters=32, input_shape=(64,64,3), kernel_size=(3,3) , padding='valid'))
model.add(Activation('relu'))
# Max Pooling
model.add(MaxPooling2D(pool_size=(2,2), padding='valid'))

# 2nd Convolutional Layer
model.add(Conv2D(filters=64, kernel_size=(3,3), padding='valid'))
model.add(Activation('relu'))
# Max Pooling
model.add(MaxPooling2D(pool_size=(2,2), padding='valid'))

# 3rd Convolutional Layer
model.add(Conv2D(filters=64, kernel_size=(3,3),  padding='valid'))
model.add(Activation('relu'))
# Max Pooling
model.add(MaxPooling2D(pool_size=(2,2), padding='valid'))

# 4th Convolutional Layer
model.add(Conv2D(filters=128, kernel_size=(3,3), padding='valid'))
model.add(Activation('relu'))
# Max Pooling
model.add(MaxPooling2D(pool_size=(2,2), padding='valid'))



# Passing it to a Fully Connected layer
model.add(Flatten())
# 1st Fully Connected Layer
model.add(Dense(128))
model.add(Activation('relu'))
# Add Dropout to prevent overfitting
model.add(Dropout(0.2))

# 2nd Fully Connected Layer
model.add(Dense(256))
model.add(Activation('relu'))
# Add Dropout
model.add(Dropout(0.2))

# 3rd Fully Connected Layer
model.add(Dense(128))
model.add(Activation('relu'))
# Add Dropout
model.add(Dropout(0.2))

# Output Layer
model.add(Dense(10))
model.add(Activation('softmax'))

model.summary()

Output: Model: "sequential_3"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_10 (Conv2D)           (None, 62, 62, 32)        896       
_________________________________________________________________
activation_67 (Activation)   (None, 62, 62, 32)        0         
_________________________________________________________________
max_pooling2d_10 (MaxPooling (None, 31, 31, 32)        0         
_________________________________________________________________
conv2d_11 (Conv2D)           (None, 29, 29, 64)        18496     
_________________________________________________________________
activation_68 (Activation)   (None, 29, 29, 64)        0         
_________________________________________________________________
max_pooling2d_11 (MaxPooling (None, 14, 14, 64)        0         
_________________________________________________________________
conv2d_12 (Conv2D)           (None, 12, 12, 64)        36928     
_________________________________________________________________
activation_69 (Activation)   (None, 12, 12, 64)        0         
_________________________________________________________________
max_pooling2d_12 (MaxPooling (None, 6, 6, 64)          0         
_________________________________________________________________
conv2d_13 (Conv2D)           (None, 4, 4, 128)         73856     
_________________________________________________________________
activation_70 (Activation)   (None, 4, 4, 128)         0         
_________________________________________________________________
max_pooling2d_13 (MaxPooling (None, 2, 2, 128)         0         
_________________________________________________________________
flatten_4 (Flatten)          (None, 512)               0         
_________________________________________________________________
dense_12 (Dense)             (None, 128)               65664     
_________________________________________________________________
activation_71 (Activation)   (None, 128)               0         
_________________________________________________________________
dropout_7 (Dropout)          (None, 128)               0         
_________________________________________________________________
dense_13 (Dense)             (None, 256)               33024     
_________________________________________________________________
activation_72 (Activation)   (None, 256)               0         
_________________________________________________________________
dropout_8 (Dropout)          (None, 256)               0         
_________________________________________________________________
dense_14 (Dense)             (None, 128)               32896     
_________________________________________________________________
activation_73 (Activation)   (None, 128)               0         
_________________________________________________________________
dropout_9 (Dropout)          (None, 128)               0         
_________________________________________________________________
dense_15 (Dense)             (None, 10)                1290      
_________________________________________________________________
activation_74 (Activation)   (None, 10)                0         
=================================================================
Total params: 263,050
Trainable params: 263,050
Non-trainable params: 0

We have made the model with 263K parameters.

Now we will compile our model by using Adam as the Optimizer and Categorical loss entropy as the loss.

model.compile(optimizer='adam',loss='categorical_crossentropy',metrics=['accuracy'])

Finally, we will train the model for 10 epochs using fit_generator.

his=model.fit_generator(train,epochs=10,shuffle=True)

Output: Epoch 1/10
37/52 [====================>.........] - ETA: 10s - loss: 0.6556 - accuracy: 0.777152/52 [==============================] - 35s 678ms/step - loss: 0.6685 - accuracy: 0.7783
Epoch 2/10
52/52 [==============================] - 35s 665ms/step - loss: 0.6527 - accuracy: 0.7881
Epoch 3/10
52/52 [==============================] - 35s 665ms/step - loss: 0.6350 - accuracy: 0.7862
Epoch 4/10
52/52 [==============================] - 35s 673ms/step - loss: 0.6144 - accuracy: 0.7942
Epoch 5/10
52/52 [==============================] - 36s 687ms/step - loss: 0.5790 - accuracy: 0.8161
Epoch 6/10
52/52 [==============================] - 35s 669ms/step - loss: 0.6082 - accuracy: 0.7978
Epoch 7/10
52/52 [==============================] - 35s 668ms/step - loss: 0.5465 - accuracy: 0.8094
Epoch 8/10
52/52 [==============================] - 35s 669ms/step - loss: 0.5960 - accuracy: 0.7996
Epoch 9/10
52/52 [==============================] - 35s 680ms/step - loss: 0.5819 - accuracy: 0.8076
Epoch 10/10
52/52 [==============================] - 36s 687ms/step - loss: 0.5272 - accuracy: 0.8270

Evaluating the Training Accuracy

print(model.evaluate(train))

Output: 52/52 [==============================] - 33s 640ms/step
        [0.1573140025138855, 0.8532277941703796]

We got 85% training accuracy which is good.

Lastly, we will predict the images on the trained model.

model.predict_classes(test,batch_size=None)

Output: array([7, 4, 8, 5, 0, 7, 7, 3, 6, 9, 6, 6, 2, 8, 9, 6, 4, 1, 1, 
        9, 6, 6,9, 1, 5, 3, 4, 9, 2, 9, 4, 7, 9, 9, 4, 4, 0, 3, 9, 4, 6,
        0, 5, 4, 4, 3, 1, 9, 9, 6, 2, 8, 2, 1, 2, 9, 7, 8, 3, 2, 1, 1, 9,
        7, 6, 2, 4, 2, 2, 4, 2, 6, 6, 3, 9, 4, 7, 7, 1, 1, 8, 9, 2, 5, 1,
        6, 2, 2, 6, 0, 8, 7, 2, 6, 9, 1, 9, 4, 9, 5, 9, 6, 0, 6, 2, 0, 9,
        2, 3, 8, 8, 1, 9, 5, 0, 1, 2, 5, 8, 1, 9, 6, 6, 5, 6, 2, 1, 7, 9, 
        9, 2, 6, 1, 0, 1, 4, 2, 2, 6, 6, 9, 9, 9, 9, 7, 6, 7, 1, 9, 7, 2, 
        0, 4, 1, 7, 4, 2, 6, 3, 1, 1, 2, 2, 2, 5, 6, 1, 6, 6, 1, 1, 9, 1, 
        6, 3, 2, 9, 9, 7, 6, 6, 1, 1, 1, 6, 6, 6])

Conclusion

In this blog, we have learned about how to make an image classification model using Keras.

The comment box is yours. Feel free to comment on any doubts or suggestions below.