forked from PeymanTahghighi/FundusDeepLearning
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Train.py
154 lines (126 loc) · 6.87 KB
/
Train.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
from Models import *
import tensorflow as tf
import Config
import pickle
import Fetcher
import cv2
import math
from matplotlib import pyplot as pp
import pydot
from tqdm.autonotebook import tqdm
tf.config.optimizer.set_jit(True)
data = Fetcher.DataFetcher(imagePath="Dataset/fundus", heightmapPath="Dataset/heightmap");
data.load();
batch_count_train = int(np.ceil(data.trainDataSize / Config.BATCH_SIZE));
batch_count_valid = int(np.ceil(data.validDataSize / Config.BATCH_SIZE));
batch_count_test = int(np.ceil(data.testDataSize / Config.BATCH_SIZE));
gen_to_disc_ratio = 3;
net = FundusNet(numBatchTrain=int(batch_count_train) * gen_to_disc_ratio);
net.generator.summary();
net.generator.load_weights(filepath="ckpt_weights1\\weights",by_name=True)
startEpoch = net.load_checkpoint();
# Define loss and accuracy per epoch
all_loss = np.zeros(batch_count_train, dtype='float32');
all_ssim = np.zeros(batch_count_train, dtype='float32');
all_lossValid = np.zeros(batch_count_valid, dtype='float32');
all_ssimValid = np.zeros(batch_count_valid, dtype='float32');
all_lossTest = np.zeros(batch_count_test, dtype='float32');
all_ssimTest = np.zeros(batch_count_test, dtype='float32');
# ----------------------------------------------------------------------------
test = False;
train = True;
if test == False:
for epoch in range(startEpoch, Config.EPOCHS):
if(train is True):
# Train step
print("\n===================================================================");
print("Epoch: ", epoch)
for g in tqdm(range(gen_to_disc_ratio)):
print("\n[INFO]Train generator step {}...".format(g));
print("Learning rate : ", net.generator_optimizer._decayed_lr(tf.float32).numpy());
for i in tqdm(tf.range(batch_count_train)):
i = tf.cast(i, tf.int64);
fundus, heightmap, size = data.fetchTrain(Config.BATCH_SIZE);
size = tf.convert_to_tensor(size, tf.int64);
loss, ssim = net.train_generator_step(fundus, heightmap,
globalStep=i + (epoch * batch_count_train),
batch_size=size)
all_loss[i] = loss;
all_ssim[i] = ssim;
epochLoss = np.mean(all_loss);
epochSsim = np.mean(all_ssim);
print("[TRAIN]epoch : {} \t loss : {} \t ssim : {}".format(epoch, epochLoss, epochSsim));
# -------------------------------------------------------------------------------
print("\n[INFO]Train discriminator step...");
print("Learning rate : ", net.generator_optimizer._decayed_lr(tf.float32).numpy());
for i in tqdm(tf.range(batch_count_train)):
i = tf.cast(i, tf.int64);
fundus, heightmap, size = data.fetchTrain(Config.BATCH_SIZE);
size = tf.convert_to_tensor(size, tf.int64);
gen = net.train_discriminator_step(fundus, heightmap,
globalStep=i + (epoch * batch_count_train))
# Validation step
print("\n[INFO]Start validation step...")
for i in tqdm(tf.range(batch_count_valid)):
i = tf.cast(i, tf.int64);
fundus, heightmap, size = data.fetchValid(Config.BATCH_SIZE);
size = tf.convert_to_tensor(size, tf.int64);
loss, accuracy = net.valid_step(fundus, heightmap,
globalStep=i + (epoch * batch_count_valid), batch_size=size)
all_lossValid[i] = loss;
all_ssimValid[i] = accuracy;
print()
epochLossValid = np.mean(all_lossValid);
epochSSIMValid = np.mean(all_ssimValid);
print("[VALID]epoch : {} \t loss : {} \t ssim : {}".format(epoch, epochLossValid, epochSSIMValid));
# -------------------------------------------------------------------------------
print("\n===================================================================");
# Debug every 5 epoch
if (epoch % 5 == 0):
print("\n[INFO]Start debugging step...");
for b in range(batch_count_train):
fundus, heightmap, size = data.fetchTrain(Config.BATCH_SIZE);
pred = net.generator(fundus, training=False);
prednp = pred.numpy() * 255;
heightmap = heightmap * 255;
fundus = fundus * 255;
a = len(prednp);
for i in range(size):
cv2.imwrite(filename='visualT/' + str((b * batch_count_train) + i) + ".png", img=prednp[i]);
cv2.imwrite(filename='visualT/' + str((b * batch_count_train) + i) + "_GT.png", img=heightmap[i]);
for b in range(batch_count_valid):
fundus, heightmap, size = data.fetchValid(Config.BATCH_SIZE);
pred = net.generator(fundus, training=False);
prednp = pred.numpy() * 255;
heightmap = heightmap * 255;
fundus = fundus *255;
a = len(prednp);
for i in range(size):
cv2.imwrite(filename='visual/' + str((b * batch_count_valid) + i) + ".png", img=prednp[i]);
cv2.imwrite(filename='visual/' + str((b * batch_count_valid) + i) + "_GT.png", img=heightmap[i]);
print("\n[INFO]Finished debugging step...");
# ------------------------------------------------------------------------------
net.save_checkpoint(step=epoch);
net.generator.save_weights(filepath="ckpt_weights2\\weights",save_format="h5");
else:
print("[INFO]Start testing step")
for b in range(50):
fundus, heightmap = data.getRandomTraining();
pred = net.generator(fundus, training=False);
prednp = pred.numpy() * 255;
heightmap = heightmap * 255;
fundus = fundus * 255;
cv2.imwrite(filename='visualTest/' + str(b) + "_Predicted.png", img=prednp[0]);
cv2.imwrite(filename='visualTest/' + str(b ) + "_GT.png", img=heightmap[0]);
cv2.imwrite(filename='visualTest/' + str(b) + "_Fundus.png", img=fundus[0]);
for b in range(50):
fundus, heightmap = data.getRandomValidation();
pred = net.generator(fundus, training=False);
prednp = pred.numpy() * 255;
heightmap = heightmap * 255;
fundus = fundus * 255;
cv2.imwrite(filename='visualTest/' + str(b+50) + "_Predicted.png", img=prednp[0]);
cv2.imwrite(filename='visualTest/' + str(b+50) + "_GT.png", img=heightmap[0]);
cv2.imwrite(filename='visualTest/' + str(b+50) + "_Fundus.png", img=fundus[0]);
print()
print("\n[INFO]End test step");