-
Notifications
You must be signed in to change notification settings - Fork 10
/
tester_Semantic3D.py
252 lines (203 loc) · 11.7 KB
/
tester_Semantic3D.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
from os import makedirs, system
from os.path import exists, join, dirname, abspath
from helper_ply import read_ply, write_ply
import tensorflow as tf
import numpy as np
import time
from sklearn.metrics import confusion_matrix
from tool import DataProcessing as DP
def log_string(out_str, log_out):
log_out.write(out_str + '\n')
log_out.flush()
print(out_str)
class ModelTester:
def __init__(self, model, dataset, restore_snap=None):
# Tensorflow Saver definition
my_vars = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES)
self.saver = tf.train.Saver(my_vars, max_to_keep=100)
# Create a session for running Ops on the Graph.
on_cpu = False
if on_cpu:
c_proto = tf.ConfigProto(device_count={'GPU': 0})
else:
c_proto = tf.ConfigProto()
c_proto.gpu_options.allow_growth = True
self.sess = tf.Session(config=c_proto)
self.sess.run(tf.global_variables_initializer())
if restore_snap is not None:
self.saver.restore(self.sess, restore_snap)
print("Model restored from " + restore_snap)
# Add a softmax operation for predictions
self.prob_logits = tf.nn.softmax(model.logits)
self.test_probs = [np.zeros((l.data.shape[0], model.config.num_classes), dtype=np.float16)
for l in dataset.input_trees['test']]
self.log_out = open('log_test_' + str(dataset.val_split) + '.txt', 'a')
def evaluate(self, model, dataset, gen_pseudo=None, num_votes=100):
# Smoothing parameter for votes
test_smooth = 0.98
# Initialise iterator with train data
self.sess.run(dataset.test_init_op)
if gen_pseudo:
# Number of points per class in validation set
val_proportions = np.zeros(model.config.num_classes, dtype=np.float32)
i = 0
for label_val in dataset.label_values:
if label_val not in dataset.ignored_labels:
val_proportions[i] = np.sum(
[np.sum(labels == label_val) for labels in dataset.input_labels['test']])
i += 1
# Test saving path
saving_path = time.strftime('results/Log_%Y-%m-%d_%H-%M-%S', time.gmtime())
test_path = join('test', saving_path.split('/')[-1])
makedirs(test_path) if not exists(test_path) else None
makedirs(join(test_path, 'predictions')) if not exists(join(test_path, 'predictions')) else None
makedirs(join(test_path, 'probs')) if not exists(join(test_path, 'probs')) else None
#####################
# Network predictions
#####################
step_id = 0
epoch_id = 0
last_min = -0.5
while last_min < num_votes:
try:
ops = (self.prob_logits,
model.labels,
model.inputs['input_inds'],
model.inputs['cloud_inds'],)
stacked_probs, stacked_labels, point_idx, cloud_idx = self.sess.run(ops, {model.is_training: False})
stacked_probs = np.reshape(stacked_probs, [model.config.val_batch_size, model.config.num_points,
model.config.num_classes])
for j in range(np.shape(stacked_probs)[0]):
probs = stacked_probs[j, :, :]
inds = point_idx[j, :]
c_i = cloud_idx[j][0]
self.test_probs[c_i][inds] = test_smooth * self.test_probs[c_i][inds] + (1 - test_smooth) * probs
step_id += 1
if not gen_pseudo:
log_string('Epoch {:3d}, step {:3d}. min possibility = {:.1f}'.format(epoch_id, step_id, np.min(
dataset.min_possibility['test'])), self.log_out)
else:
stacked_probs = np.reshape(stacked_probs, [-1, model.config.num_classes])
pred = np.argmax(stacked_probs, axis=-1)
invalid_idx = np.where(stacked_labels == 0)[0]
labels_valid = np.delete(stacked_labels, invalid_idx)
pred_valid = np.delete(pred, invalid_idx)
labels_valid = labels_valid - 1
correct = np.sum(pred_valid == labels_valid)
acc = correct / float(len(labels_valid))
print('step' + str(step_id) + ' acc:' + str(acc))
except tf.errors.OutOfRangeError:
# Save predicted cloud
new_min = np.min(dataset.min_possibility['test'])
log_string('Epoch {:3d}, end. Min possibility = {:.1f}'.format(epoch_id, new_min), self.log_out)
if last_min + 4 < new_min:
print('Saving clouds')
if gen_pseudo:
# Show vote results (On subcloud so it is not the good values here)
log_string('\nConfusion on sub clouds', self.log_out)
confusion_list = []
num_test = len(dataset.input_labels['test'])
for i_test in range(num_test):
probs = self.test_probs[i_test]
for l_ind, label_value in enumerate(dataset.label_values):
if label_value in dataset.ignored_labels:
probs = np.insert(probs, l_ind, 0, axis=1)
preds = dataset.label_values[np.argmax(probs, axis=1)].astype(np.int32)
labels = dataset.input_labels['test'][i_test]
# Confs
confusion_list += [confusion_matrix(labels, preds, dataset.label_values)]
# ==================================================== #
# Generate pseudo labels for subclouds #
# ==================================================== #
random_ratio = 0.05
trust_ratio = 0.01 / random_ratio
num_pts = len(preds)
trust_preds = np.zeros_like(preds, dtype=np.int32)
random_num = max(int(num_pts * random_ratio), 1)
random_idx = np.random.choice(num_pts, random_num, replace=False)
preds_random_selected = preds[random_idx]
probs_random_selected = probs[random_idx]
probs_random_selected_max_val = np.max(probs_random_selected, axis=1)
trust_idx_all = []
for i in range(dataset.num_classes):
ind_per_class = np.where(preds_random_selected == i)[0] # idx belongs to class
num_per_class = len(ind_per_class)
if num_per_class > 0:
trust_num = max(int(num_per_class * trust_ratio), 1)
probs_max_val_per_class = probs_random_selected_max_val[ind_per_class]
trust_pts_idx_per_class = probs_max_val_per_class.argsort()[-trust_num:][::-1]
trust_idx_per_class = ind_per_class[trust_pts_idx_per_class]
trust_idx_per_class = random_idx[trust_idx_per_class]
trust_idx_all.append(trust_idx_per_class)
trust_idx_all = np.concatenate(trust_idx_all, axis=0)
trust_preds[trust_idx_all] = preds[trust_idx_all]
print(np.sum(preds[trust_idx_all] == labels[trust_idx_all]) / len(trust_idx_all))
name = dataset.input_names['test'][i_test] + '.ply'
write_ply(join(dirname(test_path), name), [trust_preds], ['pred'])
# Regroup confusions
C = np.sum(np.stack(confusion_list), axis=0).astype(np.float32)
# Remove ignored labels from confusions
for l_ind, label_value in reversed(list(enumerate(dataset.label_values))):
if label_value in dataset.ignored_labels:
C = np.delete(C, l_ind, axis=0)
C = np.delete(C, l_ind, axis=1)
# Rescale with the right number of point per class
C *= np.expand_dims(val_proportions / (np.sum(C, axis=1) + 1e-6), 1)
# Compute IoUs
IoUs = DP.IoU_from_confusions(C)
m_IoU = np.mean(IoUs)
s = '{:5.2f} | '.format(100 * m_IoU)
for IoU in IoUs:
s += '{:5.2f} '.format(100 * IoU)
log_string(s + '\n', self.log_out)
if gen_pseudo:
return
# Update last_min
last_min = new_min
# Project predictions
print('\nReproject Vote #{:d}'.format(int(np.floor(new_min))))
t1 = time.time()
files = dataset.test_files
i_test = 0
for i, file_path in enumerate(files):
# Get file
points = self.load_evaluation_points(file_path)
points = points.astype(np.float16)
# Reproject probs
probs = np.zeros(shape=[np.shape(points)[0], 8], dtype=np.float16)
proj_index = dataset.test_proj[i_test]
probs = self.test_probs[i_test][proj_index, :]
# Insert false columns for ignored labels
probs2 = probs
for l_ind, label_value in enumerate(dataset.label_values):
if label_value in dataset.ignored_labels:
probs2 = np.insert(probs2, l_ind, 0, axis=1)
# Get the predicted labels
preds = dataset.label_values[np.argmax(probs2, axis=1)].astype(np.uint8)
# Save plys
cloud_name = file_path.split('/')[-1]
# Save ascii preds
ascii_name = join(test_path, 'predictions', dataset.ascii_files[cloud_name])
np.savetxt(ascii_name, preds, fmt='%d')
log_string(ascii_name + 'has saved', self.log_out)
i_test += 1
t2 = time.time()
print('Done in {:.1f} s\n'.format(t2 - t1))
# creat submission files
base_dir = dirname(abspath(__file__))
results_path = join(base_dir, test_path, 'predictions')
system('cd %s && zip -r %s/reduced8.zip *-reduced.labels && rm *-reduced.labels' % (
results_path, results_path))
system(
'cd %s && zip -r %s/semantic8.zip *.labels && rm *.labels' % (results_path, results_path))
import sys
sys.exit()
self.sess.run(dataset.test_init_op)
epoch_id += 1
step_id = 0
continue
return
@staticmethod
def load_evaluation_points(file_path):
data = read_ply(file_path)
return np.vstack((data['x'], data['y'], data['z'])).T