/
BS_Net_FC.py
183 lines (160 loc) · 8.17 KB
/
BS_Net_FC.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
# -*- coding: utf-8 -*-
"""
@ Description:
-------------
Band selection network with Fullly Connected Nets (aka. MLP)
-------------
@ Time : 2019/2/28 15:32
@ Author : Yaoming Cai
@ FileName: BS_Net_FC.py
@ Software: PyCharm
@ Blog ：https://github.com/AngryCai
@ Email : caiyaomxc@outlook.com
"""
import time
import numpy as np
import sys
from sklearn.linear_model import RidgeClassifier
from sklearn.model_selection import StratifiedKFold
sys.path.append('/home/caiyaom/python_codes/')
import tensorflow as tf
from sklearn.metrics import accuracy_score
from tensorflow.contrib.layers import *
from Helper import Dataset
from sklearn.neighbors import KNeighborsClassifier as KNN
from sklearn.svm import SVC
from utility import eval_band_cv
from Preprocessing import Processor
from sklearn.preprocessing import minmax_scale
class BS_Net_FC:
def __init__(self, lr, batch_size, epoch, n_selected_band):
self.lr = lr
self.batch_size = batch_size
self.epoch = epoch
self.n_selected_band = n_selected_band
tf.reset_default_graph()
tf.set_random_seed(133)
def net(self, x_input, is_training=True):
"""
:param x_input: with shape of (N, n_bands)
:param is_training:
:return:
"""
n_channel = x_input.get_shape().as_list()[-1]
input_norm = tf.layers.batch_normalization(x_input, training=is_training, name='input_norm')
# # attention module
dense_att_1 = tf.layers.dense(input_norm, 64, activation=tf.nn.relu,
kernel_initializer=tf.contrib.layers.xavier_initializer(), name='attention-1')
bn_att_1 = tf.nn.relu(tf.layers.batch_normalization(dense_att_1, training=is_training), name='BN-att-1')
bottleneck = tf.layers.dense(bn_att_1, 128, activation=tf.nn.relu,
kernel_initializer=tf.contrib.layers.xavier_initializer(),
name='bottleneck')
channel_weight = tf.layers.dense(bottleneck, n_channel, activation=tf.nn.sigmoid,
kernel_initializer=tf.contrib.layers.xavier_initializer(),
activity_regularizer=l1_regularizer(0.01), name='channel_weight')
channel_weight_ = tf.reshape(channel_weight, [-1, n_channel], name='weight_reshape')
reweight_out = channel_weight_ * input_norm
# # conv net
# Encoder
fcn_1 = tf.layers.dense(reweight_out, 64, kernel_initializer=tf.contrib.layers.xavier_initializer(),
name='fcn-1')
batch_norm_1 = tf.nn.relu(tf.layers.batch_normalization(fcn_1, training=is_training), name='BN-1')
fcn_2 = tf.layers.dense(batch_norm_1, 128, kernel_initializer=tf.contrib.layers.xavier_initializer(),
name='fcn-2')
batch_norm_2 = tf.nn.relu(tf.layers.batch_normalization(fcn_2, training=is_training), name='BN-2')
# Decoder
fcn_3 = tf.layers.dense(batch_norm_2, 256, kernel_initializer=tf.contrib.layers.xavier_initializer(),
name='fcn-3')
batch_norm_3 = tf.nn.relu(tf.layers.batch_normalization(fcn_3, training=is_training), name='BN-3')
fcn_4 = tf.layers.dense(batch_norm_3, n_channel, kernel_initializer=tf.contrib.layers.xavier_initializer(),
name='fcn-4')
output = tf.nn.sigmoid(tf.layers.batch_normalization(fcn_4, training=is_training), name='recons')
return channel_weight, output
def fit(self, X, img=None, gt=None):
n_sam, n_channel = X.shape
self.x_placehoder = tf.placeholder(shape=(None, n_channel), dtype=tf.float32)
self.is_training = tf.placeholder(tf.bool)
# self.is_fine_tuning = tf.placeholder(tf.bool)
channel_weight, output = self.net(self.x_placehoder, is_training=self.is_training)
tf.summary.histogram('channel_weight', channel_weight)
self.loss_recons = tf.losses.mean_squared_error(self.x_placehoder, output) + \
tf.losses.get_regularization_loss()
tf.summary.scalar('loss', self.loss_recons)
tf.summary.merge_all()
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
with tf.control_dependencies(update_ops):
train_op = tf.train.AdamOptimizer(learning_rate=self.lr).minimize(self.loss_recons)
saver = tf.train.Saver()
gpu_options = tf.GPUOptions(allow_growth=True) # allocate gpu memory according to model's need
sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpu_options))
sess.run(tf.global_variables_initializer())
merged = tf.summary.merge_all()
writer = tf.summary.FileWriter('logs', sess.graph)
dataset = Dataset(X, X)
loss_history = []
score_list = []
channel_weight_list = []
for i_epoch in range(self.epoch):
for batch_i in range(n_sam // self.batch_size):
x_batch, y_batch = dataset.next_batch(self.batch_size, shuffle=True)
train_op.run(feed_dict={self.x_placehoder: x_batch, self.is_training: True})
loss_reocns, channel_weight_pre, summury = sess.run([self.loss_recons, channel_weight, merged],
feed_dict={self.x_placehoder: X,
self.is_training: False})
print('epoch %s ==> loss=%s' % (i_epoch, loss_reocns))
loss_history.append(loss_reocns)
writer.add_summary(summury, i_epoch)
# if i_epoch >= 2:
channel_weight_list.append(channel_weight_pre)
if img is not None:
# score = self.eval_band(img, gt, channel_weight_, train_inx, test_idx, self.n_selected_band)
# score = self.eval_band_cv(img, gt, weight, self.n_selected_band, times=2)
mean_weight = np.mean(channel_weight_pre, axis=0)
band_indx = np.argsort(mean_weight)[::-1][:self.n_selected_band]
print('=============================')
print('SELECTED BAND: ', band_indx)
print('=============================')
x_new = img[:, :, band_indx]
n_row, n_clm, n_band = x_new.shape
img_ = minmax_scale(x_new.reshape((n_row * n_clm, n_band))).reshape((n_row, n_clm, n_band))
p = Processor()
img_correct, gt_correct = p.get_correct(img_, gt)
score = eval_band_cv(img_correct, gt_correct, times=20, test_size=0.95)
print('acc=', score)
score_list.append(score)
if i_epoch % 10 == 0:
np.savez('history-FC.npz', loss=loss_history, score=score_list, channel_weight=channel_weight_list)
np.savez('history-FC.npz', loss=loss_history, score=score_list, channel_weight=channel_weight_list)
saver.save(sess, './IndianPine-model-FC.ckpt')
'''
===================================
Demo: train model
===================================
'''
if __name__ == '__main__':
root = './Dataset/'
# root = '/home/caiyaom/HSI_Files/'
# im_, gt_ = 'SalinasA_corrected', 'SalinasA_gt'
im_, gt_ = 'Indian_pines_corrected', 'Indian_pines_gt'
# im_, gt_ = 'Pavia', 'Pavia_gt'
# im_, gt_ = 'PaviaU', 'PaviaU_gt'
# im_, gt_ = 'Salinas_corrected', 'Salinas_gt'
# im_, gt_ = 'Botswana', 'Botswana_gt'
# im_, gt_ = 'KSC', 'KSC_gt'
img_path = root + im_ + '.mat'
gt_path = root + gt_ + '.mat'
print(img_path)
p = Processor()
img, gt = p.prepare_data(img_path, gt_path)
# Img, Label = Img[:256, :, :], Label[:256, :]
n_row, n_column, n_band = img.shape
X_img = minmax_scale(img.reshape(n_row * n_column, n_band)).reshape((n_row, n_column, n_band))
X_train = np.reshape(X_img, (n_row * n_column, n_band))
print('training img shape: ', X_train.shape)
LR, BATCH_SIZE, EPOCH = 0.00002, 64, 100
N_BAND = 5
time_start = time.clock()
acnn = BS_Net_FC(LR, BATCH_SIZE, EPOCH, N_BAND)
acnn.fit(X_train, img=X_img, gt=gt)
run_time = round(time.clock() - time_start, 3)
print('running time=', run_time)