-
Notifications
You must be signed in to change notification settings - Fork 39
/
Copy pathmv-occ_8xb1_embodiedscan-occ-80class.py
174 lines (160 loc) · 7.19 KB
/
mv-occ_8xb1_embodiedscan-occ-80class.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
_base_ = ['../default_runtime.py']
n_points = 100000
# origin for multi-view scannet is set to 0.5
# -1.28~1.28 -> -0.78~1.78
point_cloud_range = [-3.2, -3.2, -0.78, 3.2, 3.2, 1.78]
prior_generator = dict(type='AlignedAnchor3DRangeGenerator',
ranges=[[-3.2, -3.2, -1.28, 3.2, 3.2, 1.28]],
rotations=[.0])
model = dict(
type='DenseFusionOccPredictor',
use_valid_mask=False,
use_xyz_feat=True,
point_cloud_range=point_cloud_range,
data_preprocessor=dict(type='Det3DDataPreprocessor',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
bgr_to_rgb=True,
pad_size_divisor=32),
backbone=dict(type='mmdet.ResNet',
depth=50,
num_stages=4,
out_indices=(0, 1, 2, 3),
frozen_stages=1,
norm_cfg=dict(type='BN', requires_grad=False),
norm_eval=True,
init_cfg=dict(type='Pretrained',
checkpoint='torchvision://resnet50'),
style='pytorch'),
neck=dict(type='mmdet.FPN',
in_channels=[256, 512, 1024, 2048],
out_channels=256,
num_outs=4),
backbone_3d=dict(type='MinkResNet', in_channels=3, depth=34),
neck_3d=dict(type='IndoorImVoxelNeck',
in_channels=256 + 512,
out_channels=128,
n_blocks=[1, 1, 1]),
bbox_head=dict(
type='ImVoxelOccHead',
volume_h=[20, 10, 5],
volume_w=[20, 10, 5],
volume_z=[8, 4, 2],
num_classes=81, # TO Be changed
in_channels=[128, 128, 128],
use_semantic=True),
prior_generator=prior_generator,
n_voxels=[40, 40, 16], # voxel_size=(.16, .16, .16)
coord_type='DEPTH',
)
dataset_type = 'EmbodiedScanDataset'
data_root = 'data'
class_names = ('floor', 'wall', 'chair', 'cabinet', 'door', 'table', 'couch',
'shelf', 'window', 'bed', 'curtain', 'desk', 'doorframe',
'plant', 'stairs', 'pillow', 'wardrobe', 'picture', 'bathtub',
'box', 'counter', 'bench', 'stand', 'rail', 'sink', 'clothes',
'mirror', 'toilet', 'refrigerator', 'lamp', 'book', 'dresser',
'stool', 'fireplace', 'tv', 'blanket', 'commode',
'washing machine', 'monitor', 'window frame', 'radiator', 'mat',
'shower', 'rack', 'towel', 'ottoman', 'column', 'blinds',
'stove', 'bar', 'pillar', 'bin', 'heater', 'clothes dryer',
'backpack', 'blackboard', 'decoration', 'roof', 'bag', 'steps',
'windowsill', 'cushion', 'carpet', 'copier', 'board',
'countertop', 'basket', 'mailbox', 'kitchen island',
'washbasin', 'bicycle', 'drawer', 'oven', 'piano',
'excercise equipment', 'beam', 'partition', 'printer',
'microwave', 'frame')
metainfo = dict(classes=class_names,
occ_classes=class_names,
box_type_3d='euler-depth')
backend_args = None
train_pipeline = [
dict(type='LoadAnnotations3D',
with_occupancy=True,
with_visible_occupancy_masks=True),
dict(type='MultiViewPipeline',
n_images=10,
transforms=[
dict(type='LoadImageFromFile', backend_args=backend_args),
dict(type='LoadDepthFromFile', backend_args=backend_args),
dict(type='ConvertRGBDToPoints', coord_type='CAMERA'),
dict(type='PointSample', num_points=n_points // 10),
dict(type='Resize', scale=(480, 480), keep_ratio=False)
]),
dict(type='AggregateMultiViewPoints', coord_type='DEPTH'),
dict(type='PointsRangeFilter', point_cloud_range=point_cloud_range),
dict(type='PointSample', num_points=n_points),
dict(type='ConstructMultiViewMasks'),
dict(
type='Pack3DDetInputs',
keys=['img', 'points', 'gt_bboxes_3d', 'gt_labels_3d', 'gt_occupancy'])
]
test_pipeline = [
dict(type='LoadAnnotations3D',
with_occupancy=True,
with_visible_occupancy_masks=True),
dict(type='MultiViewPipeline',
n_images=20,
ordered=True,
transforms=[
dict(type='LoadImageFromFile', backend_args=backend_args),
dict(type='LoadDepthFromFile', backend_args=backend_args),
dict(type='ConvertRGBDToPoints', coord_type='CAMERA'),
dict(type='PointSample', num_points=n_points // 10),
dict(type='Resize', scale=(480, 480), keep_ratio=False)
]),
dict(type='AggregateMultiViewPoints', coord_type='DEPTH'),
dict(type='PointsRangeFilter', point_cloud_range=point_cloud_range),
dict(type='PointSample', num_points=n_points),
dict(type='ConstructMultiViewMasks'),
dict(
type='Pack3DDetInputs',
keys=['img', 'points', 'gt_bboxes_3d', 'gt_labels_3d', 'gt_occupancy'])
]
train_dataloader = dict(batch_size=1,
num_workers=1,
persistent_workers=True,
sampler=dict(type='DefaultSampler', shuffle=True),
dataset=dict(type=dataset_type,
data_root=data_root,
ann_file='embodiedscan_infos_train.pkl',
pipeline=train_pipeline,
test_mode=False,
filter_empty_gt=True,
box_type_3d='Euler-Depth',
metainfo=metainfo))
val_dataloader = dict(batch_size=1,
num_workers=1,
persistent_workers=True,
drop_last=False,
sampler=dict(type='DefaultSampler', shuffle=False),
dataset=dict(type=dataset_type,
data_root=data_root,
ann_file='embodiedscan_infos_val.pkl',
pipeline=test_pipeline,
test_mode=True,
filter_empty_gt=True,
box_type_3d='Euler-Depth',
metainfo=metainfo))
test_dataloader = val_dataloader
val_evaluator = dict(type='OccupancyMetric')
test_evaluator = val_evaluator
train_cfg = dict(type='EpochBasedTrainLoop', max_epochs=24, val_interval=4)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
# optimizer
optim_wrapper = dict(type='OptimWrapper',
optimizer=dict(type='AdamW', lr=1e-4, weight_decay=0.01),
clip_grad=dict(max_norm=35., norm_type=2))
param_scheduler = dict(type='MultiStepLR',
begin=0,
end=24,
by_epoch=True,
milestones=[16, 22],
gamma=0.1)
custom_hooks = [dict(type='EmptyCacheHook', after_iter=True)]
# hooks
default_hooks = dict(
checkpoint=dict(type='CheckpointHook', interval=1, max_keep_ckpts=4))
# runtime
find_unused_parameters = True # only 1 of 4 FPN outputs is used