forked from HuangCongQing/mmdetection3d-note
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathhv_pointpillars_secfpn_ouster.py
127 lines (125 loc) · 6.09 KB
/
hv_pointpillars_secfpn_ouster.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
'''
Description:
Author: HCQ
Company(School): UCAS
Email: [email protected]
Date: 2021-09-12 11:16:43
LastEditTime: 2021-11-01 14:22:02
FilePath: /mmdetection3d/configs/_base_/models/hv_pointpillars_secfpn_ouster.py
'''
# voxel_size = [0.16, 0.16, 4] # 长宽高
voxel_size = [0.25, 0.25, 10.5] # 长宽高 z高度要和point_cloud_range高度一致z:[-3, 10]
# RuntimeError: cannot perform reduction function max on tensor with no elements because the operation does not have an identity
model = dict(
type='VoxelNet', # voxelnet.py模型名字 mmdet3d/models/detectors/__init__.py mmdet3d/models/detectors/voxelnet.py
voxel_layer=dict(
max_num_points=32, # max_points_per_voxel
# point_cloud_range=[0, -39.68, -3, 69.12, 39.68, 1], # [0, -40, -3, 60, 40, 7.5] 上层已修改
point_cloud_range = [-30, -40, -3, 80, 40, 7.5] ,
voxel_size=voxel_size,
max_voxels=(16000, 40000) # (training, testing) max_voxels
),
# 'pillar_features'torch.Size([17842, 64]) # pillar特征（C, P）的Tensor，特征维度C=64，Pillar非空P=17842个
voxel_encoder=dict(
type='PillarFeatureNet', # init--> from .pillar_encoder import PillarFeatureNet --> mmdet3d/models/voxel_encoders/pillar_encoder.py
in_channels=4,
feat_channels=[64], # pillar特征（C, P）的Tensor，特征维度C=64，Pillar非空P=17842个
with_distance=False,
voxel_size=voxel_size,
# point_cloud_range=[0, -39.68, -3, 69.12, 39.68, 1]),# 上层已修改
point_cloud_range = [-30, -40, -3, 80, 40, 7.5]
),
middle_encoder=dict(
type='PointPillarsScatter', in_channels=64, output_shape=[496, 432]), # 生成伪造图像，图像维度为(1,64,496,432) mmdet3d/models/middle_encoders/pillar_scatter.py
backbone=dict( # 调用注册器的backbone
type='SECOND', # backbone名字
# conv_cfg=dict(type='DCN', bias=False) #
in_channels=64, #
layer_nums=[3, 5, 5],
layer_strides=[2, 2, 2],
out_channels=[64, 128, 256]), # [64, 128, 256]
neck=dict(
type='SECONDFPN',
in_channels=[64, 128, 256],
upsample_strides=[1, 2, 4],
out_channels=[128, 128, 128]),
bbox_head=dict(
type='Anchor3DHead',
num_classes=7, # 修改为7类=============================
in_channels=384,
feat_channels=384,
use_direction_classifier=True,
anchor_generator=dict( # 生成anchor
type='Anchor3DRangeGenerator', # https://mmdetection3d.readthedocs.io/zh_CN/latest/api.html#mmdet3d.core.anchor.AlignedAnchor3DRangeGenerator
# ranges=[ # ？？？？？？？？？？？修改生成anchor的总范围=============================================================================
# [0, -39.68, -0.6, 70.4, 39.68, -0.6], # (x_min, y_min, z_min, x_max, y_max, z_max).
# [0, -39.68, -0.6, 70.4, 39.68, -0.6],
# [0, -39.68, -1.78, 70.4, 39.68, -1.78],
# ],
ranges=[ # 修改生成anchor的总范围=============================================================================
[-30, -40, -0.37, 60, 40, -0.37], # (x_min, y_min, z_min, x_max, y_max, z_max).
[-30, -40, -0.47, 60, 40, -0.47],
[-30, -40, -3.44, 60, 40, -3.44], # 可参考configs/_base_/models/hv_pointpillars_fpn_nus.py
],
# sizes=[[0.6, 0.8, 1.73], [0.6, 1.76, 1.73], [1.6, 3.9, 1.56]], # ['Pedestrian', 'Cyclist', 'Car'] Anchor size with shape [N, 3], in order of x, y, z.
# sizes=[[6.5, 13, 6.5], [3.5, 7, 3], [3.9, 1.6, 1.56]],
sizes=[[6.5, 13, 6.5], [3.5, 7, 3], [3.9, 1.6, 1.56]], # 3D sizes of anchors. class_names = 'Truck','Auxiliary','Car','Excavator','Widebody','Pedestrian' ('Truck','Car','Pedestrian','Excavator','Widebody','Auxiliary','Others')
rotations=[0, 1.57],
reshape_out=False),
diff_rad_by_sin=True,
# bbox_coder=dict(type='DeltaXYZWLHRBBoxCoder'),
bbox_coder=dict(type='DeltaXYZWLHRBBoxCoderOuster'), # 修改为DeltaXYZWLHRBBoxCoderOuster mmdet3d/core/bbox/coders/delta_xyzwhlr_bbox_coder_ouster.py============================
# 分类loss，回归loss，朝向loss
loss_cls=dict(
type='FocalLoss',
use_sigmoid=True,
gamma=2.0,
alpha=0.25,
loss_weight=1.0),
loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=2.0),
loss_dir=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.2)),
# model training and testing settings
train_cfg=dict(
assigner=[ # 'Truck','Auxiliary','Car'
dict( # for Pedestrian
type='MaxIoUAssigner',
iou_calculator=dict(type='BboxOverlapsNearest3D'),
pos_iou_thr=0.5,
neg_iou_thr=0.35,
min_pos_iou=0.35,
ignore_iof_thr=-1),
dict( # for Cyclist
type='MaxIoUAssigner',
iou_calculator=dict(type='BboxOverlapsNearest3D'),
pos_iou_thr=0.5,
neg_iou_thr=0.35,
min_pos_iou=0.35,
ignore_iof_thr=-1),
# dict( # for Car
# type='MaxIoUAssigner',
# iou_calculator=dict(type='BboxOverlapsNearest3D'),
# pos_iou_thr=0.6,
# neg_iou_thr=0.45,
# min_pos_iou=0.45,
# ignore_iof_thr=-1),
dict( # for Car
type='MaxIoUAssigner',
iou_calculator=dict(type='BboxOverlapsNearest3D'),
pos_iou_thr=0.45,
neg_iou_thr=0.25,
min_pos_iou=0.25,# 修改
ignore_iof_thr=-1),
],
allowed_border=0,
pos_weight=-1,
debug=False),
test_cfg=dict(
use_rotate_nms=True,
nms_across_levels=False,
nms_thr=0.01, # 去重 nms_thr=0.01,
score_thr=0.0001, # 修改阈值 score_thr=0.1,
min_bbox_size=0,
nms_pre=100,
max_num=5) # max_num
)