Handwritten character recognition with Python ¶

In this machine learning project, we will recognize handwritten characters, i.e, English alphabets from A-Z. This we are going to achieve by modeling a neural network that will have to be trained over a dataset containing images of alphabets.

First we import all necessary libraries :

In [1]:

from keras.datasets import mnist
import matplotlib.pyplot as plt
import cv2
import numpy as np
from keras.models import Sequential
from keras.layers import Dense, Flatten, Conv2D, MaxPool2D, Dropout
from keras.optimizers import SGD, Adam
from keras.callbacks import ReduceLROnPlateau, EarlyStopping
from keras.utils import to_categorical
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from keras.utils import np_utils
import matplotlib.pyplot as plt
from tqdm import tqdm_notebook
from sklearn.utils import shuffle

Using TensorFlow backend.
/home/webtunix/.local/lib/python3.5/site-packages/tensorflow/python/framework/dtypes.py:493: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint8 = np.dtype([("qint8", np.int8, 1)])
/home/webtunix/.local/lib/python3.5/site-packages/tensorflow/python/framework/dtypes.py:494: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_quint8 = np.dtype([("quint8", np.uint8, 1)])
/home/webtunix/.local/lib/python3.5/site-packages/tensorflow/python/framework/dtypes.py:495: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint16 = np.dtype([("qint16", np.int16, 1)])
/home/webtunix/.local/lib/python3.5/site-packages/tensorflow/python/framework/dtypes.py:496: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_quint16 = np.dtype([("quint16", np.uint16, 1)])
/home/webtunix/.local/lib/python3.5/site-packages/tensorflow/python/framework/dtypes.py:497: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  _np_qint32 = np.dtype([("qint32", np.int32, 1)])
/home/webtunix/.local/lib/python3.5/site-packages/tensorflow/python/framework/dtypes.py:502: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.
  np_resource = np.dtype([("resource", np.ubyte, 1)])

Read the data...

In [2]:

data = pd.read_csv('archive/A_Z Handwritten Data.csv').astype('float32')

Split data the X - Our data , and y - the prdict label

In [3]:

X = data.drop('0',axis = 1)
y = data['0']

Reshaping the data in csv file so that it can be displayed as an image...

In [4]:

train_x, test_x, train_y, test_y = train_test_split(X, y, test_size = 0.2)
train_x = np.reshape(train_x.values, (train_x.shape[0], 28,28))
test_x = np.reshape(test_x.values, (test_x.shape[0], 28,28))

print("Train data shape: ", train_x.shape)
print("Test data shape: ", test_x.shape)

Train data shape:  (297960, 28, 28)
Test data shape:  (74490, 28, 28)

Dictionary for getting characters from index values...

In [5]:

word_dict = {0:'A',1:'B',2:'C',3:'D',4:'E',5:'F',6:'G',7:'H',8:'I',9:'J',10:'K',11:'L',12:'M',13:'N',14:'O',15:'P',16:'Q',17:'R',18:'S',19:'T',20:'U',21:'V',22:'W',23:'X', 24:'Y',25:'Z'}

Plotting the number of alphabets in the dataset...

In [6]:

train_yint = np.int0(y)
count = np.zeros(26, dtype='int')
for i in train_yint:
    count[i] +=1

alphabets = []
for i in word_dict.values():
    alphabets.append(i)

fig, ax = plt.subplots(1,1, figsize=(10,10))
ax.barh(alphabets, count)

plt.xlabel("Number of elements ")
plt.ylabel("Alphabets")
plt.grid()
plt.show()

Shuffling the data ...

In [7]:

shuff = shuffle(train_x[:100])
fig, ax = plt.subplots(3,3, figsize = (10,10))
axes = ax.flatten()

for i in range(9):
    axes[i].imshow(np.reshape(shuff[i], (28,28)), cmap="Greys")
plt.show()

Reshaping the training & test dataset so that it can be put in the model...

In [8]:

train_X = train_x.reshape(train_x.shape[0],train_x.shape[1],train_x.shape[2],1)
print("New shape of train data: ", train_X.shape)

test_X = test_x.reshape(test_x.shape[0], test_x.shape[1], test_x.shape[2],1)
print("New shape of train data: ", test_X.shape)

New shape of train data:  (297960, 28, 28, 1)
New shape of train data:  (74490, 28, 28, 1)

Converting the labels to categorical values...

In [9]:

train_yOHE = to_categorical(train_y, num_classes = 26, dtype='int')
print("New shape of train labels: ", train_yOHE.shape)

test_yOHE = to_categorical(test_y, num_classes = 26, dtype='int')
print("New shape of test labels: ", test_yOHE.shape)

New shape of train labels:  (297960, 26)
New shape of test labels:  (74490, 26)

CNN model..

In [11]:

model = Sequential()

model.add(Conv2D(filters=32, kernel_size=(3, 3), activation='relu', input_shape=(28,28,1)))
model.add(MaxPool2D(pool_size=(2, 2), strides=2))

model.add(Conv2D(filters=64, kernel_size=(3, 3), activation='relu', padding = 'same'))
model.add(MaxPool2D(pool_size=(2, 2), strides=2))

model.add(Conv2D(filters=128, kernel_size=(3, 3), activation='relu', padding = 'valid'))
model.add(MaxPool2D(pool_size=(2, 2), strides=2))

model.add(Flatten())

model.add(Dense(64,activation ="relu"))
model.add(Dense(128,activation ="relu"))

model.add(Dense(26,activation ="softmax"))



model.compile(optimizer = Adam(), loss='categorical_crossentropy', metrics=['accuracy'])
reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2, patience=1, min_lr=0.0001)
early_stop = EarlyStopping(monitor='val_loss', min_delta=0, patience=2, verbose=0, mode='auto')


history = model.fit(train_X, train_yOHE, epochs=1, callbacks=[reduce_lr, early_stop],  validation_data = (test_X,test_yOHE))


model.summary()
model.save(r'model_hand.h5')

Train on 297960 samples, validate on 74490 samples
Epoch 1/1
297960/297960 [==============================] - 430s 1ms/step - loss: 1.1870 - acc: 0.8983 - val_loss: 0.0836 - val_acc: 0.9761
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_4 (Conv2D)            (None, 26, 26, 32)        320       
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 13, 13, 32)        0         
_________________________________________________________________
conv2d_5 (Conv2D)            (None, 13, 13, 64)        18496     
_________________________________________________________________
max_pooling2d_5 (MaxPooling2 (None, 6, 6, 64)          0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 4, 4, 128)         73856     
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 2, 2, 128)         0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 512)               0         
_________________________________________________________________
dense_4 (Dense)              (None, 64)                32832     
_________________________________________________________________
dense_5 (Dense)              (None, 128)               8320      
_________________________________________________________________
dense_6 (Dense)              (None, 26)                3354      
=================================================================
Total params: 137,178
Trainable params: 137,178
Non-trainable params: 0
_________________________________________________________________

Displaying the accuracies & losses for train & validation set...

In [13]:

print("The validation accuracy is :", history.history['val_acc'])
print("The training accuracy is :", history.history['acc'])
print("The validation loss is :", history.history['val_loss'])
print("The training loss is :", history.history['loss'])

The validation accuracy is : [0.9760907504363002]
The training accuracy is : [0.8983185662505034]
The validation loss is : [0.08358431425431195]
The training loss is : [1.1870133685692619]

Making model predictions...

In [14]:

pred = model.predict(test_X[:9])
print(test_X.shape)

(74490, 28, 28, 1)

Displaying some of the test images & their predicted labels...

In [15]:

fig, axes = plt.subplots(3,3, figsize=(8,9))
axes = axes.flatten()

for i,ax in enumerate(axes):
    img = np.reshape(test_X[i], (28,28))
    ax.imshow(img, cmap="Greys")
    pred = word_dict[np.argmax(test_yOHE[i])]
    ax.set_title("Prediction: "+pred)
    ax.grid()