-
Notifications
You must be signed in to change notification settings - Fork 128
/
utils.py
312 lines (255 loc) · 11.7 KB
/
utils.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
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
"""
Utilities for for OmniDet.
# author: Varun Ravi Kumar <[email protected]>
Parts of the code adapted from https://github.com/nianticlabs/monodepth2
Please refer to the license of the above repo.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; Authors provide no warranty with the software
and are not liable for anything.
"""
import os
import time
import numpy as np
import torch
from colorama import Fore, Style
from ruamel import yaml
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms
class TrainUtils:
def __init__(self, args):
"""Train Utils class providing training utilities for distance, semantic and motion estimation
:param args: input params from config file
"""
self.args = args
self.device = args.device
self.log_path = os.path.join(args.output_directory, args.model_name)
assert args.input_height % 32 == 0, "'height' must be a multiple of 32"
assert args.input_width % 32 == 0, "'width' must be a multiple of 32"
self.models = dict()
self.parameters_to_train = []
self.epoch = 0
self.step = 0
self.start_time = time.time()
self.trans_pil = transforms.ToPILImage()
self.optimizer = None
self.lr_scheduler = None
self.writers = dict()
for mode in ["train", "val"]:
self.writers[mode] = SummaryWriter(os.path.join(self.log_path, mode))
def inputs_to_device(self, inputs):
for key, ipt in inputs.items():
inputs[key] = ipt.to(self.device)
def set_train(self):
"""Convert all models to training mode"""
for m in self.models.values():
m.train()
def set_eval(self):
"""Convert all models to testing/evaluation mode"""
for m in self.models.values():
m.eval()
def log_time(self, batch_idx, duration, loss, data_time, gpu_time):
"""Print a logging statement to the terminal"""
samples_per_sec = self.args.batch_size / duration
time_sofar = time.time() - self.start_time
training_time_left = (self.num_total_steps / self.step - 1.0) * time_sofar if self.step > 0 else 0
print(f"{Fore.GREEN}epoch {self.epoch:>3}{Style.RESET_ALL} "
f"| batch {batch_idx:>6} "
f"| current lr {self.optimizer.param_groups[0]['lr']:.4f} "
f"| examples/s: {samples_per_sec:5.1f} "
f"| {Fore.RED}loss: {loss:.5f}{Style.RESET_ALL} "
f"| {Fore.BLUE}time elapsed: {self.sec_to_hm_str(time_sofar)}{Style.RESET_ALL} "
f"| {Fore.CYAN}time left: {self.sec_to_hm_str(training_time_left)}{Style.RESET_ALL} "
f"| CPU/GPU time: {data_time:0.1f}s/{gpu_time:0.1f}s")
def configure_optimizers(self):
self.optimizer = torch.optim.Adam(self.parameters_to_train, self.args.learning_rate)
self.lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(self.optimizer, self.args.scheduler_step_size)
def save_model(self):
"""Save model weights to disk"""
save_folder = os.path.join(self.log_path, "models", f"weights_{self.epoch}", str(self.step))
if not os.path.exists(save_folder):
os.makedirs(save_folder)
for model_name, model in self.models.items():
save_path = os.path.join(save_folder, f"{model_name}.pth")
to_save = model.state_dict()
if model_name == 'encoder':
# save the sizes - these are needed at prediction time
to_save['height'] = self.args.input_height
to_save['width'] = self.args.input_width
torch.save(to_save, save_path)
save_path = os.path.join(save_folder, "adam.pth")
if self.epoch > 50: # Optimizer file is quite large! Sometimes, life is a compromise.
torch.save(self.optimizer.state_dict(), save_path)
def load_model(self):
"""Load model(s) from disk"""
self.args.pretrained_weights = os.path.expanduser(self.args.pretrained_weights)
assert os.path.isdir(self.args.pretrained_weights), f"Cannot find folder {self.args.pretrained_weights}"
print(f"=> Loading model from folder {self.args.pretrained_weights}")
for n in self.args.models_to_load:
print(f"Loading {n} weights...")
path = os.path.join(self.args.pretrained_weights, f"{n}.pth")
model_dict = self.models[n].state_dict()
pretrained_dict = torch.load(path, map_location=self.args.device)
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
model_dict.update(pretrained_dict)
self.models[n].load_state_dict(model_dict)
# loading optimizer state
if not self.args.freeze_encoder:
optimizer_load_path = os.path.join(self.args.pretrained_weights, f"{self.args.optimizer}.pth")
if os.path.isfile(optimizer_load_path):
print(f"Loading {self.args.optimizer} weights")
optimizer_dict = torch.load(optimizer_load_path, map_location=self.args.device)
self.optimizer.load_state_dict(optimizer_dict)
else:
print(f"Cannot find {self.args.optimizer} weights so {self.args.optimizer} is randomly initialized")
def save_args(self):
"""Save arguments to disk so we know what we ran this experiment with"""
models_dir = os.path.join(self.log_path, "models")
if not os.path.exists(models_dir):
os.makedirs(models_dir)
to_save = self.args.copy()
with open(os.path.join(models_dir, 'params.yaml'), 'w') as f:
yaml.dump(to_save, f)
def sec_to_hm(self, t):
"""Convert time in seconds to time in hours, minutes and seconds
e.g. 10239 -> (2, 50, 39)
"""
t = int(t)
s = t % 60
t //= 60
m = t % 60
t //= 60
return t, m, s
def sec_to_hm_str(self, t):
"""Convert time in seconds to a nice string
e.g. 10239 -> '02h50m39s'
"""
h, m, s = self.sec_to_hm(t)
return f"{h:02d}h{m:02d}m{s:02d}s"
class AverageMeter:
"""Computes and stores the average and current value"""
def __init__(self):
self.initialized = False
self.val = None
self.avg = None
self.sum = None
self.count = None
def initialize(self, val, weight):
self.val = val
self.avg = val
self.sum = val * weight
self.count = weight
self.initialized = True
def update(self, val, weight=1):
if not self.initialized:
self.initialize(val, weight)
else:
self.add(val, weight)
def add(self, val, weight):
self.val = val
self.sum += val * weight
self.count += weight
self.avg = self.sum / self.count
def value(self):
return self.val
def average(self):
return self.avg
@staticmethod
def accuracy(preds, label):
valid = (label >= 0)
acc_sum = (valid * (preds == label)).sum()
valid_sum = valid.sum()
acc = float(acc_sum) / (valid_sum + 1e-10)
return acc, valid_sum
class Tupperware(dict):
MARKER = object()
def __init__(self, value=None):
if value is None:
pass
elif isinstance(value, dict):
for key in value:
self.__setitem__(key, value[key])
else:
raise TypeError('expected dict')
def __setitem__(self, key, value):
if isinstance(value, dict) and not isinstance(value, Tupperware):
value = Tupperware(value)
super(Tupperware, self).__setitem__(key, value)
def __getitem__(self, key):
found = self.get(key, Tupperware.MARKER)
if found is Tupperware.MARKER:
found = Tupperware()
super(Tupperware, self).__setitem__(key, found)
return found
__setattr__, __getattr__ = __setitem__, __getitem__
class IoU:
"""Computes the intersection over union (IoU) per class and corresponding mean (mIoU).
The predictions are first accumulated in a confusion matrix and the IoU is computed from it as follows:
IoU = true_positive / (true_positive + false_positive + false_negative).
:param num_classes (int): number of classes in the classification problem
:param dataset (string): woodscape_raw
:param ignore_index (int or iterable, optional): Index of the classes to ignore when computing the IoU.
"""
def __init__(self, num_classes, dataset, ignore_index=None):
super().__init__()
self.conf_metric = np.ndarray((num_classes, num_classes), dtype=np.int32)
self.num_classes = num_classes
self.dataset = dataset
self.classes = dict(woodscape_raw=["void", "road", "lanemarks", "curb", "person",
"rider", "vehicles", "bicycle", "motorcycle""traffic_sign"],
motion=['static', 'motion'], )
self.reset()
if ignore_index is None:
self.ignore_index = None
elif isinstance(ignore_index, int):
self.ignore_index = (ignore_index,)
else:
try:
self.ignore_index = tuple(ignore_index)
except TypeError:
raise ValueError("'ignore_index' must be an int or iterable")
def reset(self):
self.conf_metric.fill(0)
def add(self, predicted, target):
"""Adds the predicted and target pair to the IoU metric."""
predicted = predicted.view(-1).cpu().numpy()
target = target.view(-1).cpu().numpy()
# hack for bin counting 2 arrays together
x = predicted + self.num_classes * target
bincount_2d = np.bincount(x.astype(np.int32), minlength=self.num_classes ** 2)
assert bincount_2d.size == self.num_classes ** 2
conf = bincount_2d.reshape((self.num_classes, self.num_classes))
self.conf_metric += conf
def value(self):
"""Computes the IoU and mean IoU.
The mean computation ignores NaN elements of the IoU array.
Returns: Tuple: (class_iou, mIoU). The first output is the per class IoU, for K classes it's numpy.ndarray with
K elements. The second output, is the mean IoU.
"""
if self.ignore_index is not None:
for index in self.ignore_index:
self.conf_metric[:, self.ignore_index] = 0
self.conf_metric[self.ignore_index, :] = 0
true_positive = np.diag(self.conf_metric)
false_positive = np.sum(self.conf_metric, 0) - true_positive
false_negative = np.sum(self.conf_metric, 1) - true_positive
# Just in case we get a division by 0, ignore/hide the error
with np.errstate(divide='ignore', invalid='ignore'):
iou = true_positive / (true_positive + false_positive + false_negative)
class_dict = self.classes[self.dataset]
class_iou = dict(zip(class_dict, iou))
return class_iou, np.nanmean(iou)
def semantic_color_encoding(args):
semantic_classes = dict(void=(0, 0, 0),
road=(149, 213, 0),
lanemarks=(216, 45, 128),
curb=(0, 140, 88),
person=(255, 0, 0),
rider=(255, 255, 255),
vehicles=(0, 0, 255),
bicycle=(0, 255, 255),
motorcycle=(30, 170, 250),
traffic_sign=(0, 128, 128))
color_encoding = np.zeros((args.semantic_num_classes, 3), dtype=np.uint8)
for i, (k, v) in enumerate(semantic_classes.items()):
color_encoding[i] = v
return color_encoding