forked from TarekIdrees/ASL-using-CNN
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathASL_CNN.py
230 lines (168 loc) · 6.97 KB
/
ASL_CNN.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
# -*- coding: utf-8 -*-
"""ASL_CNN.ipynb
Automatically generated by Colaboratory.
Original file is located at
https://colab.research.google.com/drive/1M5JA5_OV0N_biYsciMnmlUvrAor8Yx_u
**Set up colab for ASL data from kaggel**
"""
! pip install kaggle
! mkdir ~/.kaggle
! cp kaggle.json ~/.kaggle/
! chmod 600 ~/.kaggle/kaggle.json
! kaggle datasets download grassknoted/asl-alphabet
! unzip asl-alphabet.zip
"""**Import needed libraries**"""
import numpy as np
import os
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split, GridSearchCV
import cv2
import keras
from keras.layers import Conv2D, MaxPool2D, Flatten, Dense, Dropout, BatchNormalization,Activation,MaxPooling2D
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras import regularizers
from sklearn.metrics import accuracy_score
from sklearn.metrics import precision_score
from sklearn.metrics import recall_score
train_dir = 'asl_alphabet_train//asl_alphabet_train'
test_dir = 'asl_alphabet_test//asl_alphabet_test'
labels_dict = {'A': 0, 'B': 1, 'C': 2, 'D': 3, 'E': 4, 'F': 5, 'G': 6, 'H': 7, 'I': 8, 'J': 9, 'K': 10, 'L': 11,
'M': 12,
'N': 13, 'O': 14, 'P': 15, 'Q': 16, 'R': 17, 'S': 18, 'T': 19, 'U': 20, 'V': 21, 'W': 22, 'X': 23,
'Y': 24,
'Z': 25, 'space': 26, 'del': 27, 'nothing': 28}
def load_train_data():
Y_train = []
X_train = []
size = 64, 64
#number_of_images_per_folder = 0
images_per_folder = 0
print("LOADING DATA FROM : ", end="")
for folder in os.listdir(train_dir):
print(folder, end=' | ')
for image in os.listdir(train_dir + "/" + folder):
if images_per_folder == 2000:
images_per_folder = 0
break
# read image
temp_img = cv2.imread(train_dir + '/' + folder + '/' + image)
# resize image
temp_img = cv2.resize(temp_img, size)
#load converted classes
Y_train.append(labels_dict[folder])
X_train.append(temp_img)
images_per_folder = images_per_folder +1
#convert X_train to numpy
X_train = np.array(X_train)
#normalize the pixels of X_train
X_train = X_train.astype('float32')/255.0
#convert Y_train to numpy
Y_train = np.array(Y_train)
print()
print('Loaded', len(X_train), 'images for training,', 'Train data shape =', X_train.shape)
return X_train, Y_train
def load_test_data():
labels = []
X_test = []
size = 64, 64
for image in os.listdir(test_dir):
# read image
temp_img = cv2.imread(test_dir + '/'+ image)
# resize image
temp_img = cv2.resize(temp_img, size)
# load converted classes
labels.append(labels_dict[image.split('_')[0]])
X_test.append(temp_img)
#convert X_test to numpy
X_test = np.array(X_test)
#normalize pixels of X_test
X_test = X_test.astype('float32')/255.0
#convert Y_test to numpy
Y_test = np.array(labels)
print('Loaded', len(X_test), 'images for testing,', 'Test data shape =', X_test.shape)
return X_test, Y_test
"""**Load data with Gray**"""
X_train, Y_train = load_train_data()
X_test, Y_test = load_test_data()
"""**Create first CNN model**
"""
def create_model():
model = Sequential()
model.add(Conv2D(16, kernel_size = [3,3], padding = 'same', activation = 'relu', input_shape = (64,64,3)))
model.add(Conv2D(32, kernel_size = [3,3], padding = 'same', activation = 'relu'))
model.add(MaxPool2D(pool_size = [3,3]))
model.add(Conv2D(32, kernel_size = [3,3], padding = 'same', activation = 'relu'))
model.add(Conv2D(64, kernel_size = [3,3], padding = 'same', activation = 'relu'))
model.add(MaxPool2D(pool_size = [3,3]))
model.add(Conv2D(128, kernel_size = [3,3], padding = 'same', activation = 'relu', input_shape = (64,64,3)))
model.add(Conv2D(256, kernel_size = [3,3], padding = 'same', activation = 'relu'))
model.add(MaxPool2D(pool_size = [3,3]))
model.add(BatchNormalization())
model.add(Flatten())
model.add(Dropout(rate = 0.5))
model.add(Dense(512, activation = 'relu',kernel_regularizer = regularizers.l2(0.001)))
model.add(Dense(29, activation = 'softmax'))
model.summary()
return model
"""**Crete second CNN model**"""
def create_model_2():
model = Sequential()
model.add(Conv2D(32, (5, 5), input_shape=(64, 64, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D((2, 2)))
model.add(Conv2D(64, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D((2, 2)))
model.add(Conv2D(64, (3, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D((2, 2)))
model.add(BatchNormalization())
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dense(29, activation='softmax'))
model.summary()
return model
"""**Description of model 2**"""
model_2 = create_model_2()
"""**Description of model 1**"""
model = create_model()
"""**Compile fisrt model**"""
model.compile(optimizer='adam',loss = 'sparse_categorical_crossentropy',metrics=['accuracy'])
"""**Compile second model**"""
model_2.compile(optimizer='adam',loss = 'sparse_categorical_crossentropy',metrics=['accuracy'])
"""**Fit first model with 5 epochs**"""
epochs = 5
ThisModel = model.fit(X_train,Y_train,epochs = epochs, batch_size=64,verbose = 1)
"""**Fit second model with 5 epochs**"""
ThisModel_2 = model_2.fit(X_train, Y_train,epochs=5,batch_size=64,verbose=1)
"""**Report accuracy of model 1**"""
modelLoss , modelAccuracy = model.evaluate(X_test,Y_test)
print('Test Loss is %d'%modelLoss)
print('Test Accuracy is %d'%modelAccuracy)
"""**Report accuracy of model 2**"""
modelLoss_2 , modelAccuracy_2 = model_2.evaluate(X_test,Y_test)
print('Test Loss is %d'%modelLoss_2)
print('Test Accuracy is %d'%modelAccuracy_2)
"""**Predict using first model**"""
Y_predict=model.predict(X_test)
classes_y=np.argmax(Y_predict,axis=1)
"""**Predict using second model**"""
Y_predict_2=model_2.predict(X_test)
classes_y_2=np.argmax(Y_predict_2,axis=1)
"""**Report accuracy,recall and precission**"""
# calculate accuracy
accuracy = accuracy_score(Y_test, classes_y)
print('Model accuracy is: ', accuracy)
# Model Precision: what percentage of positive tuples are labeled as such?
print("Precision:", precision_score(Y_test, classes_y, average='micro'))
# Model Recall: what percentage of positive tuples are labelled as such?
print("Recall:", recall_score(Y_test, classes_y, average='micro'))
"""**Report accuracy,recall and precission**"""
# calculate accuracy
accuracy_2 = accuracy_score(Y_test, classes_y_2)
print('Model accuracy is: ', accuracy_2)
# Model Precision: what percentage of positive tuples are labeled as such?
print("Precision:", precision_score(Y_test, classes_y_2, average='micro'))
# Model Recall: what percentage of positive tuples are labelled as such?
print("Recall:", recall_score(Y_test, classes_y_2, average='micro'))