-
Notifications
You must be signed in to change notification settings - Fork 5
/
cma_segformer_robotcar.yaml
191 lines (191 loc) · 5.19 KB
/
cma_segformer_robotcar.yaml
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
data:
class_path: data_modules.CombinedDataModule
init_args:
batch_size: 1
num_workers: 4
load_config:
train:
RobotCar:
load_pseudo_labels: True
pseudo_label_dir: pseudo_labels_train_RobotCar_cma_segformer
load_keys:
- image
- semantic
- image_ref
transforms:
- class_path: data_modules.transforms.ToTensor
- class_path: data_modules.transforms.RandomCrop
init_args:
size:
- 1024
- 1024
cat_max_ratio: 0.75
- class_path: data_modules.transforms.RandomHorizontalFlip
- class_path: data_modules.transforms.ConvertImageDtype
- class_path: data_modules.transforms.Normalize
val:
RobotCar:
load_keys:
- image
- semantic
transforms:
- class_path: data_modules.transforms.ToTensor
- class_path: data_modules.transforms.ConvertImageDtype
- class_path: data_modules.transforms.Normalize
test:
RobotCar:
load_keys:
- image
- semantic
transforms:
- class_path: data_modules.transforms.ToTensor
- class_path: data_modules.transforms.ConvertImageDtype
- class_path: data_modules.transforms.Normalize
# ACG:
# conditions:
# - fog
# - night
# - rain
# - snow
# load_keys:
# - image
# - semantic
# transforms:
# - class_path: data_modules.transforms.ToTensor
# - class_path: data_modules.transforms.ConvertImageDtype
# - class_path: data_modules.transforms.Normalize
predict:
RobotCar:
predict_on: val
load_keys:
- image
transforms:
- class_path: data_modules.transforms.ToTensor
- class_path: data_modules.transforms.ConvertImageDtype
- class_path: data_modules.transforms.Normalize
model:
class_path: models.CMAModel
init_args:
entropy_loss_weight: 0.01
contrastive_loss_weight: 1.0
freeze_decoder: True
projection_head_lr_factor: 10.0
ema_momentum: 0.9999
backbone:
class_path: models.backbones.MixVisionTransformer
init_args:
model_type: mit_b5
pretrained: cityscapes
head:
class_path: models.heads.SegFormerHead
init_args:
in_channels:
- 64
- 128
- 320
- 512
in_index:
- 0
- 1
- 2
- 3
channels: 768
num_classes: 19
input_transform: multiple_select
pretrained: cityscapes
contrastive_head:
class_path: models.heads.ProjectionHead
init_args:
in_channels:
- 64
- 128
- 320
- 512
in_index:
- 0
- 1
- 2
- 3
channels: 128
input_transform: resize_concat
alignment_backbone:
class_path: models.backbones.VGG
init_args:
model_type: vgg16
pretrained: imagenet
out_indices:
- 2
- 3
- 4
alignment_head:
class_path: models.heads.UAWarpCHead
init_args:
in_index:
- 0
- 1
input_transform: multiple_select
estimate_uncertainty: True
pretrained: megadepth
self_training_loss:
class_path: torch.nn.CrossEntropyLoss
init_args:
ignore_index: 255
entropy_loss:
class_path: models.losses.NormalizedEntropyLoss
contrastive_loss:
class_path: models.losses.CDCLoss
init_args:
feat_dim: 128
temperature: 0.3
num_grid: 7
queue_len: 65536
warm_up_steps: 2500
confidence_threshold: 0.2
metrics:
val:
RobotCar:
- class_path: helpers.metrics.IoU
init_args:
num_classes: 19
ignore_index: 255
average: macro
over_present_classes: True
test:
RobotCar:
- class_path: helpers.metrics.IoU
init_args:
num_classes: 19
ignore_index: 255
average: macro
over_present_classes: True
# ACG:
# - class_path: helpers.metrics.IoU
# init_args:
# num_classes: 19
# ignore_index: 255
# average: macro
optimizer:
class_path: torch.optim.AdamW
init_args:
lr: 0.00001
weight_decay: 0.01
lr_scheduler:
class_path: helpers.lr_scheduler.LinearWarmupLinearLR
init_args:
warmup_iters: 1500
warmup_ratio: 0.000001
max_steps: 10000
trainer:
max_steps: 10000
val_check_interval: 1000
check_val_every_n_epoch: null
sync_batchnorm: True
logger:
class_path: pytorch_lightning.loggers.CSVLogger
init_args:
save_dir: logs/cma_segformer_robotcar
callbacks:
- class_path: pytorch_lightning.callbacks.LearningRateMonitor
- class_path: pytorch_lightning.callbacks.ModelCheckpoint
init_args:
every_n_train_steps: 1000