-
Notifications
You must be signed in to change notification settings - Fork 2
/
segment.py
242 lines (203 loc) · 10.8 KB
/
segment.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
import argparse
import os
import h5py
import numpy
import cupy
from keras import optimizers
from keras.utils import to_categorical
from keras.models import load_model
from keras.utils.io_utils import HDF5Matrix
import threading
import utility
import process
import models
class threadsafe_iter:
"""Takes an iterator/generator and makes it thread-safe by
serializing call to the `next` method of given iterator/generator.
"""
def __init__(self, it):
self.it = it
self.lock = threading.Lock()
def __iter__(self):
return self
def __next__(self):
with self.lock:
return self.it.__next__()
def threadsafe_generator(f):
"""A decorator that takes a generator function and makes it thread-safe.
"""
def g(*a, **kw):
return threadsafe_iter(f(*a, **kw))
return g
if __name__ == '__main__':
# mode argument
parser = argparse.ArgumentParser()
parser.add_argument('--epochs', type=int, default=1000, help='Number of epochs')
parser.add_argument('--batch_size', type=int, default=16, help='Size of batch')
parser.add_argument('--dim', type=int, default=256, help='ELM weight size')
parser.add_argument('--lr', type=float, default=0.0005, help='Learning rate')
parser.add_argument('--margin', type=float, default=0.4, help='Radius of sampling sphere')
parser.add_argument('--train', type=bool, default=True, help='Conduct training')
parser.add_argument('--normals', type=bool, default=True, help='Use normals for training')
parser.add_argument('--sample_size', type=int, default=1024, help='Number of sampling points for distance field calculation')
args = parser.parse_args()
batch_size = args.batch_size
epochs = args.epochs
DATA_DIR = #your directory
SAVE_DIR = #your directory
num_classes = 50
train_inst = 12135
test_inst = 2873
#check if data preparation is necessary
data_list_train = utility.makeList(os.path.join(DATA_DIR, 'train_files.txt'))
data_list_test = utility.makeList(os.path.join(DATA_DIR, 'test_files.txt'))
#prepare samplings points
basis_name = 'rands' + str(args.dim) + '.h5'
point_name = 'points' + str(args.sample_size) + 'dim' + str(args.dim) + 'margin' + str(args.margin) + '.h5'
utility.prepSamplePoints(args.sample_size, os.path.join(SAVE_DIR, point_name), os.path.join(SAVE_DIR, basis_name), args.dim, args.margin)
for num in range(len(data_list_train)):
original_name = data_list_train[num] + '.h5'
svd_name = data_list_train[num] + 'margin' + str(args.margin) + 'svd.h5'
weight_name = data_list_train[num] + 'margin' + str(args.margin) + 'dim' + str(args.dim) + '.h5'
#calculate distances
print('---- train data ' + str(num) + ' ----')
if not os.path.isfile(os.path.join(SAVE_DIR, svd_name)):
print('**** Calculating Distance ****')
process.calcDistField(os.path.join(SAVE_DIR, point_name), os.path.join(DATA_DIR, original_name), os.path.join(SAVE_DIR, svd_name))
#convert them into ELM weights
if not os.path.isfile(os.path.join(SAVE_DIR, weight_name)):
print('**** Processing ELM ****')
process.saveELM(os.path.join(SAVE_DIR,svd_name), os.path.join(DATA_DIR, original_name), os.path.join(SAVE_DIR, weight_name), os.path.join(SAVE_DIR, point_name), os.path.join(SAVE_DIR, basis_name), args.dim)
for num in range(len(data_list_test)):
original_name = data_list_test[num] + '.h5'
svd_name = data_list_test[num]+ 'margin' + str(args.margin) + 'svd.h5'
weight_name = data_list_test[num]+ 'margin' + str(args.margin) + 'dim' + str(args.dim) + '.h5'
#calculate distances
print('---- test data ' + str(num) + ' ----')
if not os.path.isfile(os.path.join(SAVE_DIR, svd_name)):
print('**** Calculating Distance ****')
process.calcDistField(os.path.join(SAVE_DIR, point_name), os.path.join(DATA_DIR, original_name), os.path.join(SAVE_DIR, svd_name))
#convert them into ELM weights
if not os.path.isfile(os.path.join(SAVE_DIR, weight_name)):
print('**** Processing ELM ****')
process.saveELM(os.path.join(SAVE_DIR,svd_name), os.path.join(DATA_DIR, original_name), os.path.join(SAVE_DIR, weight_name), os.path.join(SAVE_DIR, point_name), os.path.join(SAVE_DIR, basis_name), args.dim)
if args.train:
cat_file = os.path.join(DATA_DIR, 'synsetoffset2category.txt')
cat_dict = {}
with open(cat_file, 'r') as f:
for line in f:
ls = line.strip().split()
cat_dict[ls[0]] = ls[1]
cat_dict = {k:v for k,v in cat_dict.items()}
seg_classes = {'Earphone': [16, 17, 18], 'Motorbike': [30, 31, 32, 33, 34, 35], 'Rocket': [41, 42, 43], 'Car': [8, 9, 10, 11], 'Laptop': [28, 29], 'Cap': [6, 7], 'Skateboard': [44, 45, 46], 'Mug': [36, 37], 'Guitar': [19, 20, 21], 'Bag': [4, 5], 'Lamp': [24, 25, 26, 27], 'Table': [47, 48, 49], 'Airplane': [0, 1, 2, 3], 'Pistol': [38, 39, 40], 'Chair': [12, 13, 14, 15], 'Knife': [22, 23]}
for cat in sorted(seg_classes.keys()):
print(cat, seg_classes[cat])
classes = dict(zip(cat_dict, range(len(cat_dict))))
print(classes)
shape_ious = {cat:[] for cat in seg_classes.keys()}
seg_label_to_cat = {} # {0:Airplane, 1:Airplane, ...49:Table}
for cat in seg_classes.keys():
for label in seg_classes[cat]:
seg_label_to_cat[label] = cat
@threadsafe_generator
def generator(data_list, batch_size):
while True:
numlab = numpy.arange(len(data_list))
numpy.random.shuffle(numlab)
for curnum in range(len(data_list)):
num = numlab[curnum]
hdf5_name = data_list[num]+ 'margin' + str(args.margin) + 'dim' + str(args.dim) + '.h5'
hdf5_file = os.path.join(SAVE_DIR, hdf5_name)
hdf5_name2 = data_list[num]+ '.h5'
hdf5_file2 = os.path.join(DATA_DIR, hdf5_name2)
wgt = HDF5Matrix(hdf5_file, 'data')
if args.normals:
pts = HDF5Matrix(hdf5_file2, 'data')
nms = HDF5Matrix(hdf5_file2, 'normal')
size = wgt.end
y = HDF5Matrix(hdf5_file2, 'segment')
batchnm = int(numpy.ceil(size/batch_size))
blab = numpy.arange(batchnm)
numpy.random.shuffle(blab)
for itt in range(batchnm):
idx = blab[itt] * batch_size
if idx + batch_size >= size:
end = size
else:
end = idx + batch_size
x1 = wgt[idx:end]
if args.normals:
x2 = pts[idx:end]
x3 = nms[idx:end]
xbatch = [x1, x2, x3]
else:
xbatch = x1
yield xbatch, to_categorical(y[idx:end],num_classes=num_classes)
max_acc = 0
max_iou = 0
max_shapeious = None
#prepare model
overalldim = args.dim
model_name = 'model' + str(overalldim) + 'segment.h5'
if args.normals:
mlp = models.defineModelSegmentPN(2048, args.dim, num_classes)
adam = optimizers.Adam(lr=args.lr, beta_1=0.99, beta_2=0.999, epsilon=1e-08, decay=0.000)
mlp.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])
else:
mlp = models.defineModelSegment(2048, args.dim, num_classes)
adam = optimizers.Adam(lr=args.lr, beta_1=0.99, beta_2=0.999, epsilon=1e-08, decay=0.000)
mlp.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])
# for multi gpu
train_generator = generator(data_list_train, args.batch_size)
test_generator = generator(data_list_test, args.batch_size)
#iterate
#training process
for it in range(args.epochs):
pred_val = numpy.zeros((test_inst,2048,num_classes))
test_labels = numpy.zeros((test_inst,2048,num_classes))
print('**** Epoch %03d ****' % (it))
mlp.fit_generator(
epochs=1,
generator=train_generator,
steps_per_epoch=int(numpy.ceil(train_inst / args.batch_size)),
max_queue_size=10,
workers=1,
use_multiprocessing=False,
shuffle=False,verbose=1)
curid = 0
for i in range(int(numpy.ceil(test_inst/args.batch_size))):
x, y = next(test_generator)
batchend = y.shape[0]
pred_val[curid:curid+batchend] = mlp.predict(x)
test_labels[curid:curid+batchend] = y
curid += batchend
shape_ious = {cat:[] for cat in seg_classes.keys()}
for i in range(test_inst):
curpred = pred_val[i]
segp = numpy.argmax(curpred,axis=1)
segl = numpy.argmax(test_labels[i,:],axis=1).flatten()
cat = seg_label_to_cat[segl[0]]
part_ious = [0.0 for _ in range(len(seg_classes[cat]))]
for l in seg_classes[cat]:
if (numpy.sum(segl==l) == 0) and (numpy.sum(segp==l) == 0): # part is not present, no prediction as well
part_ious[l-seg_classes[cat][0]] = 1.0
else:
part_ious[l-seg_classes[cat][0]] = numpy.sum((segl==l) & (segp==l)) / float(numpy.sum((segl==l) | (segp==l)))
shape_ious[cat].append(numpy.mean(part_ious))
all_shape_ious = []
for cat in shape_ious.keys():
for iou in shape_ious[cat]:
all_shape_ious.append(iou)
shape_ious[cat] = numpy.mean(shape_ious[cat])
mean_shape_ious = numpy.mean(numpy.fromiter(shape_ious.values(),dtype=float))
for cat in sorted(shape_ious.keys()):
print('eval mIoU of %s:\t %f' %(cat, shape_ious[cat]))
print('eval mean mIoU: %f' % (mean_shape_ious))
print('eval mean mIoU (all shapes): %f' % (numpy.mean(all_shape_ious)))
if max_iou < numpy.mean(all_shape_ious):
max_iou = numpy.mean(all_shape_ious)
max_shapeious = shape_ious
print('Maximum IoU: %f' % max_iou)
if max_shapeious:
for cat in sorted(max_shapeious.keys()):
print('eval mIoU of %s:\t %f' %(cat, max_shapeious[cat]))