-
Notifications
You must be signed in to change notification settings - Fork 0
/
Chawan_06_01.py
385 lines (307 loc) · 17.7 KB
/
Chawan_06_01.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
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
# Chawan, Varsha Rani
# 1001-553-524
# 2018-12-10
# Assignment-06-01
import sys
if sys.version_info[0] < 3:
import Tkinter as tk
else:
import tkinter as tk
import matplotlib
matplotlib.use('TkAgg')
from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg
import numpy as np
import tensorflow as tf
import Chawan_06_02
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sn
class MainWindow(tk.Tk):
"""
This class creates and controls the main window frames and widgets
Chawan Varsha Rani
"""
def __init__(self, debug_print_flag=False):
tk.Tk.__init__(self)
self.debug_print_flag = debug_print_flag
self.master_frame = tk.Frame(self)
self.master_frame.grid(row=0, column=0, sticky=tk.N + tk.E + tk.S + tk.W)
# set the properties of the row and columns in the master frame
self.rowconfigure(0, weight=1, minsize=500)
self.columnconfigure(0, weight=1, minsize=500)
self.master_frame.rowconfigure(2, weight=10, minsize=100, uniform='xx')
self.master_frame.columnconfigure(0, weight=1, minsize=200, uniform='xx')
# Create an object for plotting graphs in the left frame
self.left_frame = tk.Frame(self.master_frame)
self.left_frame.grid(row=2, column=0, sticky=tk.N + tk.E + tk.S + tk.W)
self.display_error = CNN(self, self.left_frame, debug_print_flag=self.debug_print_flag)
class CNN:
"""
This class creates and controls the sliders , buttons , drop down in the frame which
are used to display decision boundary and generate samples and train .
"""
def __init__(self, root, master, debug_print_flag=False):
self.master = master
self.root = root
#########################################################################
# Set up the constants and default values
#########################################################################
self.alpha = 0.1
self.lamda = 0.01
self.height = 32
self.width = 32
self.channel = 3
self.n_inputs = self.height * self.width
self.F1 = 32
self.K1 = 3
self.conv1_stride = 1
self.conv1_pad = "SAME"
self.F2 = 32
self.K2 = 3
self.conv2_pad = "SAME"
self.conv2_stride = 1
self.F3 = 32
self.K3 = 3
self.conv3_stride = 1
self.conv3_pad = "SAME"
self.pool_map = self.F1
self.pool2_map = self.F2
self.pool3_map = self.F3
self.n_fullyconnected = 32
self.n_output = 10
self.train_samples = 20
self.total_samples = int(50000 * (self.train_samples / 100))
self.x_train, self.y_train, one_hot_train = Chawan_06_02.load_training_data()
self.x_train = self.x_train[:self.total_samples]
self.y_train = self.y_train[:self.total_samples]
self.x_test, self.y_test, one_hot_test = Chawan_06_02.load_test_data()
self.epoch = 0
self.epocs_list = []
self.accu_list = []
tf.reset_default_graph()
self.X = tf.placeholder(tf.float32, shape=[None, self.height, self.width, self.channel], name="X")
self.Y = tf.placeholder(tf.int32, shape=[None])
self.current_iterations = 0
self.batch_size = 100
self.regularizer = tf.contrib.layers.l2_regularizer(scale=self.lamda)
self.conv1 = tf.layers.conv2d(self.X, filters=self.F1, kernel_size=self.K1, strides=self.conv1_stride,
padding=self.conv1_pad, activation=tf.nn.relu,kernel_regularizer=self.regularizer, name="conv1")
self.pool1 = tf.nn.max_pool(self.conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
self.conv2 = tf.layers.conv2d(self.pool1, filters=self.F2, kernel_size=self.K2, strides=self.conv2_stride,
padding=self.conv2_pad, activation=tf.nn.relu,kernel_regularizer=self.regularizer, name="conv2")
self.pool2 = tf.nn.max_pool(self.conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
self.conv3 = tf.layers.conv2d(self.pool2, filters=self.F3, kernel_size=self.K3, strides=self.conv3_stride,
padding=self.conv3_pad, activation=tf.nn.relu,kernel_regularizer=self.regularizer, name="conv3")
self.pool3 = tf.nn.max_pool(self.conv3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
shape_list = self.pool3.get_shape().as_list()
self.pool3_flat = tf.reshape(self.pool3, shape=[-1, shape_list[1] * shape_list[2] * shape_list[3]])
self.fullyconn = tf.layers.dense(self.pool3_flat, self.n_fullyconnected, activation=None,kernel_regularizer=self.regularizer, name="fc1")
self.logits = tf.layers.dense(self.fullyconn, self.n_output, name="output")
self.sess = tf.Session()
self.init = tf.global_variables_initializer()
self.sess.run(self.init)
#########################################################################
# Set up the plotting frame and controls frame
#########################################################################
master.rowconfigure(0, weight=10, minsize=200)
master.columnconfigure(0, weight=1)
self.plot_frame = tk.Frame(self.master, borderwidth=10, relief=tk.SUNKEN)
self.plot_frame.grid(row=0, column=0, columnspan=1, sticky=tk.N + tk.E + tk.S + tk.W)
self.figure, self.axes_mul = plt.subplots(1, 2, figsize=(16, 6))
self.axes = self.figure.gca()
self.canvas = FigureCanvasTkAgg(self.figure, master=self.plot_frame)
self.plot_widget = self.canvas.get_tk_widget().pack(side=tk.TOP, fill=tk.BOTH, expand=True)
self.controls_frame = tk.Frame(self.master)
self.controls_frame.grid(row=1, column=0, sticky=tk.N + tk.E + tk.S + tk.W)
#########################################################################
# Set up the control widgets such as sliders ,buttons and dropdowns
#########################################################################
self.alpha_slider = tk.Scale(self.controls_frame, variable=tk.DoubleVar(), orient=tk.HORIZONTAL,
from_=0.000, to_=1.0, resolution=0.001, bg="#DDDDDD",
activebackground="#FF0000", highlightcolor="#00FFFF", label="Learning Rate",
command=lambda event: self.alpha_slider_callback())
self.alpha_slider.set(self.alpha)
self.alpha_slider.bind("<ButtonRelease-1>", lambda event: self.alpha_slider_callback())
self.alpha_slider.grid(row=0, column=0, sticky=tk.N + tk.E + tk.S + tk.W)
self.lamda_slider = tk.Scale(self.controls_frame, variable=tk.DoubleVar(), orient=tk.HORIZONTAL,
from_=0.0, to_=1.0, resolution=0.01, bg="#DDDDDD",
activebackground="#FF0000", highlightcolor="#00FFFF", label="Lambda",
command=lambda event: self.lamda_slider_callback())
self.lamda_slider.set(self.lamda)
self.lamda_slider.bind("<ButtonRelease-1>", lambda event: self.lamda_slider_callback())
self.lamda_slider.grid(row=0, column=1, sticky=tk.N + tk.E + tk.S + tk.W)
self.F1_slider = tk.Scale(self.controls_frame, variable=tk.IntVar(), orient=tk.HORIZONTAL,
from_=1, to_=64, resolution=1, bg="#DDDDDD",
activebackground="#FF0000", highlightcolor="#00FFFF",
label="F1",
command=lambda event: self.F1_slider_callback())
self.F1_slider.set(self.F1)
self.F1_slider.bind("<ButtonRelease-1>", lambda event: self.F1_slider_callback())
self.F1_slider.grid(row=0, column=2, sticky=tk.N + tk.E + tk.S + tk.W)
self.F2_slider = tk.Scale(self.controls_frame, variable=tk.IntVar(), orient=tk.HORIZONTAL,
from_=1, to_=64, resolution=1, bg="#DDDDDD",
activebackground="#FF0000", highlightcolor="#00FFFF",
label="F2",
command=lambda event: self.F2_slider_callback())
self.F2_slider.set(self.F2)
self.F2_slider.bind("<ButtonRelease-1>", lambda event: self.F2_slider_callback())
self.F2_slider.grid(row=0, column=3, sticky=tk.N + tk.E + tk.S + tk.W)
self.K1_slider = tk.Scale(self.controls_frame, variable=tk.IntVar(), orient=tk.HORIZONTAL,
from_=3, to_=7, resolution=2, bg="#DDDDDD",
activebackground="#FF0000", highlightcolor="#00FFFF",
label="K1",
command=lambda event: self.K1_slider_callback())
self.K1_slider.set(self.K1)
self.K1_slider.bind("<ButtonRelease-1>", lambda event: self.K1_slider_callback())
self.K1_slider.grid(row=1, column=2, sticky=tk.N + tk.E + tk.S + tk.W)
self.K2_slider = tk.Scale(self.controls_frame, variable=tk.IntVar(), orient=tk.HORIZONTAL,
from_=3, to_=7, resolution=2, bg="#DDDDDD",
activebackground="#FF0000", highlightcolor="#00FFFF",
label="K2",
command=lambda event: self.K2_slider_callback())
self.K2_slider.set(self.K2)
self.K2_slider.bind("<ButtonRelease-1>", lambda event: self.K2_slider_callback())
self.K2_slider.grid(row=1, column=3, sticky=tk.N + tk.E + tk.S + tk.W)
self.train_samples_slider = tk.Scale(self.controls_frame, variable=tk.IntVar(), orient=tk.HORIZONTAL,
from_=0, to_=100, resolution=1, bg="#DDDDDD",
activebackground="#FF0000", highlightcolor="#00FFFF",
label="Train Samples",
command=lambda event: self.train_samples_slider_callback())
self.train_samples_slider.set(self.train_samples)
self.train_samples_slider.bind("<ButtonRelease-1>", lambda event: self.train_samples_slider_callback())
self.train_samples_slider.grid(row=0, column=4, sticky=tk.N + tk.E + tk.S + tk.W)
self.adjust_weight_button = tk.Button(self.controls_frame, text="Train", width=16,
command=self.adjust_weight_button_callback)
self.adjust_weight_button.grid(row=1, column=0)
self.reset_weight_button = tk.Button(self.controls_frame, text="Reset Weights", width=16,
command=self.reset_weight_button_callback)
self.reset_weight_button.grid(row=1, column=1)
def plot_ErrorRate(self, epoc,acc,cm):
#########################################################################
# Freeze the weights and calculate o/p , find the index and plot
#########################################################################
self.axes.cla()
self.axes.xaxis.set_visible(True)
self.axes_mul[0].cla()
self.axes_mul[0].plot(epoc, acc, marker='o', markersize=5)
self.axes_mul[0].set_title("CNN")
self.axes_mul[0].set_xlabel('Epochs')
self.axes_mul[0].set_ylabel('Percentage Error')
self.axes_mul[1].cla()
self.axes_mul[1].set_xlabel("Predicted")
self.axes_mul[1].set_ylabel("True")
plt.title("Confusion Matrix ")
df_cm = pd.DataFrame(cm,range(10),range(10))
sn.set(font_scale=1.4)
sn.heatmap(df_cm,annot=True,cbar=False,fmt='g')
plt.tight_layout()
self.canvas.draw()
def get_next_batch(self, batch_size):
self.current_iterations = 0
start_index = (self.current_iterations * batch_size) % len(self.y_train)
end_index = start_index + batch_size
x_batch = self.x_train[start_index:end_index]
y_batch = self.y_train[start_index:end_index]
self.current_iterations = self.current_iterations + 1
return x_batch, y_batch
def train_model(self):
self.epoch += 1
Xentropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=self.logits, labels=self.Y)
loss = tf.reduce_mean(Xentropy)
l2_loss = tf.losses.get_regularization_loss()
loss += l2_loss
optimizer = tf.train.GradientDescentOptimizer(self.alpha)
# optimizer = tf.train.AdamOptimizer(self.alpha)
training_op = optimizer.minimize(loss)
num_exam = len(self.y_train)
for iteration in range(num_exam // self.batch_size):
X_batch, Y_batch = self.get_next_batch(self.batch_size)
self.sess.run(training_op, feed_dict={self.X: X_batch, self.Y: Y_batch})
self.test_data()
def reset_tensor_variables(self):
self.epoch = 0
self.epocs_list = []
self.accu_list = []
tf.reset_default_graph()
self.X = tf.placeholder(tf.float32, shape=[None, self.height, self.width, self.channel], name="X")
self.Y = tf.placeholder(tf.int32, shape=[None])
self.current_iterations = 0
self.batch_size = 100
self.regularizer = tf.contrib.layers.l2_regularizer(scale=self.lamda)
self.conv1 = tf.layers.conv2d(self.X, filters=self.F1, kernel_size=self.K1, strides=self.conv1_stride,
padding=self.conv1_pad, activation=tf.nn.relu,
kernel_regularizer=self.regularizer, name="conv1")
self.pool1 = tf.nn.max_pool(self.conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
self.conv2 = tf.layers.conv2d(self.pool1, filters=self.F2, kernel_size=self.K2, strides=self.conv2_stride,
padding=self.conv2_pad, activation=tf.nn.relu,
kernel_regularizer=self.regularizer, name="conv2")
self.pool2 = tf.nn.max_pool(self.conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
self.conv3 = tf.layers.conv2d(self.pool2, filters=self.F3, kernel_size=self.K3, strides=self.conv3_stride,
padding=self.conv3_pad, activation=tf.nn.relu,
kernel_regularizer=self.regularizer, name="conv3")
self.pool3 = tf.nn.max_pool(self.conv3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
shape_list = self.pool3.get_shape().as_list()
self.pool3_flat = tf.reshape(self.pool3, shape=[-1, shape_list[1] * shape_list[2] * shape_list[3]])
self.fullyconn = tf.layers.dense(self.pool3_flat, self.n_fullyconnected, activation=None,
kernel_regularizer=self.regularizer, name="fc1")
self.logits = tf.layers.dense(self.fullyconn, self.n_output, name="output")
self.sess = tf.Session()
self.init = tf.global_variables_initializer()
self.sess.run(self.init)
def test_data(self):
correct = tf.nn.in_top_k(self.logits, self.Y, 1)
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))
accuracypercent = accuracy*100
error = 100.00 - accuracypercent
acc_test = error.eval(session=self.sess, feed_dict={self.X: self.x_test, self.Y: self.y_test})
self.epocs_list.append(self.epoch)
self.accu_list.append(acc_test)
_, top_predicted_label = tf.nn.top_k(self.logits, k=1, sorted=False)
top_predicted_label = tf.squeeze(top_predicted_label, axis=1)
cm = tf.confusion_matrix(
self.y_test, top_predicted_label, num_classes=self.n_output)
cmeval = cm.eval(session=self.sess, feed_dict={self.X: self.x_test, self.Y: self.y_test})
self.plot_ErrorRate(self.epocs_list, self.accu_list, cmeval)
def alpha_slider_callback(self):
self.alpha = np.float(self.alpha_slider.get())
def lamda_slider_callback(self):
self.lamda = np.float(self.lamda_slider.get())
def F1_slider_callback(self):
self.F1 = self.F1_slider.get()
self.reset_tensor_variables()
self.test_data()
def F2_slider_callback(self):
self.F2 = self.F2_slider.get()
self.reset_tensor_variables()
self.test_data()
def K1_slider_callback(self):
self.K1 = self.K1_slider.get()
self.reset_tensor_variables()
self.test_data()
def K2_slider_callback(self):
self.K2 = self.K2_slider.get()
self.reset_tensor_variables()
self.test_data()
def train_samples_slider_callback(self):
self.train_samples = self.train_samples_slider.get()
self.total_samples = int(50000 * (self.train_samples / 100))
self.x_train, self.y_train, one_hot_train = Chawan_06_02.load_training_data()
self.x_train = self.x_train[:self.total_samples]
self.y_train = self.y_train[:self.total_samples]
def adjust_weight_button_callback(self): #train
self.train_model()
def reset_weight_button_callback(self):
self.reset_tensor_variables()
self.test_data()
#########################################################################
# Logic to close Main Window
#########################################################################
def close_window_callback(root):
if tk.messagebox.askokcancel("Quit", "Do you really wish to quit?"):
root.destroy()
main_window = MainWindow(debug_print_flag=False)
main_window.wm_state('zoomed')
main_window.title('Assignment_06 -- Chawan')
main_window.minsize(800, 600)
main_window.protocol("WM_DELETE_WINDOW", lambda root_window=main_window: close_window_callback(root_window))
main_window.mainloop()