s3dis-seg.py 4.55 KB
Newer Older
ZCMax's avatar
ZCMax committed
1
# For S3DIS seg we usually do 13-class segmentation
2
3
class_names = ('ceiling', 'floor', 'wall', 'beam', 'column', 'window', 'door',
               'table', 'chair', 'sofa', 'bookcase', 'board', 'clutter')
4
metainfo = dict(classes=class_names)
ZCMax's avatar
ZCMax committed
5
6
7
8
9
10
11
12
dataset_type = 'S3DISSegDataset'
data_root = 'data/s3dis/'
input_modality = dict(use_lidar=True, use_camera=False)
data_prefix = dict(
    pts='points',
    pts_instance_mask='instance_mask',
    pts_semantic_mask='semantic_mask')

13
14
15
16
17
18
19
20
file_client_args = dict(backend='disk')
# Uncomment the following if use ceph or other file clients.
# See https://mmcv.readthedocs.io/en/latest/api.html#mmcv.fileio.FileClient
# for more details.
# file_client_args = dict(
#     backend='petrel',
#     path_mapping=dict({
#         './data/s3dis/':
21
#         's3://s3dis/',
22
#     }))
ZCMax's avatar
ZCMax committed
23

24
25
26
27
28
29
30
31
32
33
num_points = 4096
train_area = [1, 2, 3, 4, 6]
test_area = 5
train_pipeline = [
    dict(
        type='LoadPointsFromFile',
        coord_type='DEPTH',
        shift_height=False,
        use_color=True,
        load_dim=6,
34
        use_dim=[0, 1, 2, 3, 4, 5]),
35
36
37
38
39
    dict(
        type='LoadAnnotations3D',
        with_bbox_3d=False,
        with_label_3d=False,
        with_mask_3d=False,
40
        with_seg_3d=True),
ZCMax's avatar
ZCMax committed
41
    dict(type='PointSegClassMapping'),
42
43
44
45
46
    dict(
        type='IndoorPatchPointSample',
        num_points=num_points,
        block_size=1.0,
        ignore_index=len(class_names),
47
48
49
        use_normalized_coord=True,
        enlarge_size=0.2,
        min_unique_num=None),
50
    dict(type='NormalizePointsColor', color_mean=None),
ZCMax's avatar
ZCMax committed
51
    dict(type='Pack3DDetInputs', keys=['points', 'pts_semantic_mask'])
52
53
54
55
56
57
58
59
]
test_pipeline = [
    dict(
        type='LoadPointsFromFile',
        coord_type='DEPTH',
        shift_height=False,
        use_color=True,
        load_dim=6,
60
        use_dim=[0, 1, 2, 3, 4, 5]),
ChaimZhu's avatar
ChaimZhu committed
61
62
63
64
65
66
    dict(
        type='LoadAnnotations3D',
        with_bbox_3d=False,
        with_label_3d=False,
        with_mask_3d=False,
        with_seg_3d=True),
67
    dict(type='NormalizePointsColor', color_mean=None),
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
    dict(
        # a wrapper in order to successfully call test function
        # actually we don't perform test-time-aug
        type='MultiScaleFlipAug3D',
        img_scale=(1333, 800),
        pts_scale_ratio=1,
        flip=False,
        transforms=[
            dict(
                type='GlobalRotScaleTrans',
                rot_range=[0, 0],
                scale_ratio_range=[1., 1.],
                translation_std=[0, 0, 0]),
            dict(
                type='RandomFlip3D',
                sync_2d=False,
                flip_ratio_bev_horizontal=0.0,
                flip_ratio_bev_vertical=0.0),
ZCMax's avatar
ZCMax committed
86
87
        ]),
    dict(type='Pack3DDetInputs', keys=['points'])
88
89
90
91
92
93
94
95
96
97
98
]
# construct a pipeline for data and gt loading in show function
# please keep its loading function consistent with test_pipeline (e.g. client)
# we need to load gt seg_mask!
eval_pipeline = [
    dict(
        type='LoadPointsFromFile',
        coord_type='DEPTH',
        shift_height=False,
        use_color=True,
        load_dim=6,
99
        use_dim=[0, 1, 2, 3, 4, 5]),
ZCMax's avatar
ZCMax committed
100
101
    dict(type='NormalizePointsColor', color_mean=None),
    dict(type='Pack3DDetInputs', keys=['points'])
102
103
]

ZCMax's avatar
ZCMax committed
104
105
106
107
108
109
110
111
# train on area 1, 2, 3, 4, 6
# test on area 5
train_dataloader = dict(
    batch_size=8,
    num_workers=4,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
112
113
        type=dataset_type,
        data_root=data_root,
ChaimZhu's avatar
ChaimZhu committed
114
        ann_files=[f's3dis_infos_Area_{i}.pkl' for i in train_area],
ZCMax's avatar
ZCMax committed
115
116
        metainfo=metainfo,
        data_prefix=data_prefix,
117
        pipeline=train_pipeline,
ZCMax's avatar
ZCMax committed
118
        modality=input_modality,
119
120
        ignore_index=len(class_names),
        scene_idxs=[
ChaimZhu's avatar
ChaimZhu committed
121
            f'seg_info/Area_{i}_resampled_scene_idxs.npy' for i in train_area
122
        ],
ZCMax's avatar
ZCMax committed
123
124
125
126
127
128
129
130
        test_mode=False))
test_dataloader = dict(
    batch_size=1,
    num_workers=1,
    persistent_workers=True,
    drop_last=False,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
131
132
        type=dataset_type,
        data_root=data_root,
ChaimZhu's avatar
ChaimZhu committed
133
        ann_files=f's3dis_infos_Area_{test_area}.pkl',
ZCMax's avatar
ZCMax committed
134
135
        metainfo=metainfo,
        data_prefix=data_prefix,
136
        pipeline=test_pipeline,
ZCMax's avatar
ZCMax committed
137
        modality=input_modality,
138
        ignore_index=len(class_names),
ChaimZhu's avatar
ChaimZhu committed
139
        scene_idxs=f'seg_info/Area_{test_area}_resampled_scene_idxs.npy',
ZCMax's avatar
ZCMax committed
140
141
        test_mode=True))
val_dataloader = test_dataloader
142

ZCMax's avatar
ZCMax committed
143
144
val_evaluator = dict(type='SegMetric')
test_evaluator = val_evaluator
145
146
147
148

vis_backends = [dict(type='LocalVisBackend')]
visualizer = dict(
    type='Det3DLocalVisualizer', vis_backends=vis_backends, name='visualizer')