-
Notifications
You must be signed in to change notification settings - Fork 13
/
mnist_spatial.py
199 lines (154 loc) · 6 KB
/
mnist_spatial.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
from __future__ import print_function
from __future__ import division
from six.moves import xrange
import os
import time
import tensorflow as tf
import numpy as np
from sklearn.preprocessing import StandardScaler
from lib.datasets import MNIST as Data
from lib.model import Model as BaseModel
from lib.segmentation import segmentation_adjacency, extract_features_fixed
# from lib.segmentation import slic_fixed
from lib.segmentation import quickshift_fixed
from lib.layer import SpatialCNN as Conv, FC
from lib.graph import receptive_fields, fill_features
from lib.pipeline import PreprocessedDataset, FileQueue
# SLIC_FEATURES = [4, 5, 6, 7, 8, 18, 20, 21, 22]
QUICKSHIFT_FEATURES = [4, 6, 7, 8, 24, 28, 29, 31, 37]
DATA_DIR = 'data/mnist'
# PREPROCESS_FIRST = 'data/mnist/slic_spatial'
PREPROCESS_FIRST = 'data/mnist/quickshift_spatial'
NODE_SIZE = 25
NODE_STRIDE = 4
DELTA = 3
NEIGHBORHOOD_SIZE = 25
CONNECTIVITY = 8
LEARNING_RATE = 0.001
TRAIN_DIR = None
# LOG_DIR = 'data/summaries/mnist_slic_spatial'
LOG_DIR = 'data/summaries/mnist_quickshift_spatial'
SAVE_STEP = 250
AUGMENT_TRAIN_EXAMPLES = False
DROPOUT = 0.5
BATCH_SIZE = 64
MAX_STEPS = 15000
DISPLAY_STEP = 10
# FORM_FEATURES = SLIC_FEATURES
FORM_FEATURES = QUICKSHIFT_FEATURES
NUM_FEATURES = len(FORM_FEATURES) + 1
data = Data(DATA_DIR)
# segmentation_algorithm = slic_fixed(
# num_segments=100, compactness=5, max_iterations=10, sigma=0)
segmentation_algorithm = quickshift_fixed(
ratio=1, kernel_size=2, max_dist=2, sigma=0)
feature_extraction_algorithm = extract_features_fixed(FORM_FEATURES)
def preprocess_spatial_fixed(
segmentation_algorithm, feature_extraction_algorithm, node_size,
node_stride, delta, neighborhood_size, connectivity):
def _preprocess(image):
segmentation = segmentation_algorithm(image)
adj, points, _ = segmentation_adjacency(segmentation, connectivity)
features = feature_extraction_algorithm(segmentation, image)
StandardScaler(copy=False).fit_transform(features)
fields = receptive_fields(points, adj, node_size, node_stride,
neighborhood_size, delta)
return fill_features(fields, features)
return _preprocess
preprocess_algorithm = preprocess_spatial_fixed(
segmentation_algorithm, feature_extraction_algorithm, NODE_SIZE,
NODE_STRIDE, DELTA, NEIGHBORHOOD_SIZE, CONNECTIVITY)
# Generate preprocessed dataset.
data.train = PreprocessedDataset(
os.path.join(PREPROCESS_FIRST, 'train'), data.train, preprocess_algorithm)
data.val = PreprocessedDataset(
os.path.join(PREPROCESS_FIRST, 'val'), data.val, preprocess_algorithm)
data.test = PreprocessedDataset(
os.path.join(PREPROCESS_FIRST, 'test'), data.test, preprocess_algorithm)
capacity = 10 * BATCH_SIZE
train_queue = FileQueue(data.train, BATCH_SIZE, capacity, shuffle=True)
val_queue = FileQueue(data.val, BATCH_SIZE, capacity, shuffle=True)
test_queue = FileQueue(data.test, BATCH_SIZE, capacity, shuffle=False)
placeholders = {
'features':
tf.placeholder(tf.float32,
[None, NODE_SIZE, NEIGHBORHOOD_SIZE,
NUM_FEATURES], 'features'),
'labels':
tf.placeholder(tf.uint8, [None, data.num_classes], 'labels'),
'dropout':
tf.placeholder(tf.float32, [], 'dropout'),
}
class Model(BaseModel):
def _build(self):
conv_1 = Conv(
NUM_FEATURES, 64, NEIGHBORHOOD_SIZE, logging=self.logging)
fc_1 = FC(NODE_SIZE * 64, 1024, logging=self.logging)
fc_2 = FC(
1024,
data.num_classes,
act=lambda x: x,
bias=False,
dropout=self.placeholders['dropout'],
logging=self.logging)
self.layers = [conv_1, fc_1, fc_2]
model = Model(
placeholders=placeholders,
learning_rate=LEARNING_RATE,
train_dir=TRAIN_DIR,
log_dir=LOG_DIR)
model.build()
global_step = model.initialize()
def feed_dict_with_batch(batch, dropout=0):
features = np.array([data[0] for data in batch], np.float32)
labels = np.array([data[1] for data in batch], np.uint8)
return {
placeholders['features']: features,
placeholders['labels']: labels,
placeholders['dropout']: DROPOUT,
}
try:
for step in xrange(global_step, MAX_STEPS):
t_pre = time.process_time()
batch = train_queue.dequeue()
feed_dict = feed_dict_with_batch(batch, DROPOUT)
t_pre = time.process_time() - t_pre
t_train = model.train(feed_dict, step)
if step % DISPLAY_STEP == 0:
# Evaluate on training and validation set with zero dropout.
feed_dict.update({model.placeholders['dropout']: 0})
train_info = model.evaluate(feed_dict, step, 'train')
batch = val_queue.dequeue()
feed_dict = feed_dict_with_batch(batch, DROPOUT)
val_info = model.evaluate(feed_dict, step, 'val')
log = 'step={}, '.format(step)
log += 'time={:.2f}s + {:.2f}s, '.format(t_pre, t_train)
log += 'train_loss={:.5f}, '.format(train_info[0])
log += 'train_acc={:.5f}, '.format(train_info[1])
log += 'val_loss={:.5f}, '.format(val_info[0])
log += 'val_acc={:.5f}'.format(val_info[1])
print(log)
if step % SAVE_STEP == 0:
model.save()
except KeyboardInterrupt:
print()
train_queue.close()
val_queue.close()
print('Optimization finished!')
print('Evaluate on test set. This can take a few minutes.')
try:
num_steps = data.test.num_examples // BATCH_SIZE
test_info = [0, 0]
for i in xrange(num_steps):
batch = test_queue.dequeue()
feed_dict = feed_dict_with_batch(batch, DROPOUT)
batch_info = model.evaluate(feed_dict)
test_info = [a + b for a, b in zip(test_info, batch_info)]
log = 'Test results: '
log += 'loss={:.5f}, '.format(test_info[0] / num_steps)
log += 'acc={:.5f}, '.format(test_info[1] / num_steps)
print(log)
except KeyboardInterrupt:
print()
print('Test evaluation aborted.')
test_queue.close()