Voxelnet/config.py at master · steph1793/Voxelnet · GitHub

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
from easydict import EasyDict as edict

__cfg__ = edict()

# for dataset dir
__cfg__.DATA_DIR = 'DATA_DIR/T_DATA'
__cfg__.KITTY_EVAL_SCRIPT = "kitti_eval/launch_test.sh"
__cfg__.CALIB_DIR = ''

# selected object
__cfg__.DETECT_OBJECT = 'Car'  # Pedestrian/Cyclist
__cfg__.NUM_ANCHORS_PER_CELL = 2

if __cfg__.DETECT_OBJECT == 'Car':
    __cfg__.MAX_POINT_NUMBER = 35
    __cfg__.Z_MIN = -3
    __cfg__.Z_MAX = 1
    __cfg__.Y_MIN = -40
    __cfg__.Y_MAX = 40
    __cfg__.X_MIN = 0
    __cfg__.X_MAX = 70.4
    __cfg__.VOXEL_X_SIZE = 0.2
    __cfg__.VOXEL_Y_SIZE = 0.2
    __cfg__.VOXEL_Z_SIZE = 0.4
    __cfg__.VOXEL_POINT_COUNT = 35
    __cfg__.INPUT_WIDTH = int((__cfg__.X_MAX - __cfg__.X_MIN) / __cfg__.VOXEL_X_SIZE)
    __cfg__.INPUT_HEIGHT = int((__cfg__.Y_MAX - __cfg__.Y_MIN) / __cfg__.VOXEL_Y_SIZE)
    __cfg__.INPUT_DEPTH = int((__cfg__.Z_MAX - __cfg__.Z_MIN) / __cfg__.VOXEL_Z_SIZE)
    __cfg__.LIDAR_COORD = [0, 40, 3]
    __cfg__.FEATURE_RATIO = 2
    __cfg__.FEATURE_WIDTH = int(__cfg__.INPUT_WIDTH / __cfg__.FEATURE_RATIO)
    __cfg__.FEATURE_HEIGHT = int(__cfg__.INPUT_HEIGHT / __cfg__.FEATURE_RATIO)
else:
    __cfg__.MAX_POINT_NUMBER = 45
    __cfg__.Z_MIN = -3
    __cfg__.Z_MAX = 1
    __cfg__.Y_MIN = -20
    __cfg__.Y_MAX = 20
    __cfg__.X_MIN = 0
    __cfg__.X_MAX = 48
    __cfg__.VOXEL_X_SIZE = 0.2
    __cfg__.VOXEL_Y_SIZE = 0.2
    __cfg__.VOXEL_POINT_COUNT = 45
    __cfg__.INPUT_WIDTH = int((__cfg__.X_MAX - __cfg__.X_MIN) / __cfg__.VOXEL_X_SIZE)
    __cfg__.INPUT_HEIGHT = int((__cfg__.Y_MAX - __cfg__.Y_MIN) / __cfg__.VOXEL_Y_SIZE)
    __cfg__.INPUT_DEPTH = int((__cfg__.Z_MAX - __cfg__.Z_MIN) / __cfg__.VOXEL_Z_SIZE)
    __cfg__.FEATURE_RATIO = 2
    __cfg__.LIDAR_COORD = [0, 20, 3]
    __cfg__.FEATURE_WIDTH = int(__cfg__.INPUT_WIDTH / __cfg__.FEATURE_RATIO)
    __cfg__.FEATURE_HEIGHT = int(__cfg__.INPUT_HEIGHT / __cfg__.FEATURE_RATIO)


__cfg__.SCENE_SIZE = [__cfg__.Z_MAX - __cfg__.Z_MIN, __cfg__.Y_MAX- __cfg__.Y_MIN, __cfg__.X_MAX - __cfg__.X_MIN]
__cfg__.VOXEL_SIZE = [__cfg__.VOXEL_Z_SIZE, __cfg__.VOXEL_Y_SIZE, __cfg__.VOXEL_X_SIZE]
__cfg__.GRID_SIZE = [int(A/B) for A,B in zip(__cfg__.SCENE_SIZE, __cfg__.VOXEL_SIZE)]
__cfg__.MAP_SHAPE = [__cfg__.FEATURE_HEIGHT, __cfg__.FEATURE_WIDTH]

__cfg__.IMG_WIDTH = 1242
__cfg__.IMG_HEIGHT = 375
__cfg__.IMG_CHANNEL = 3


# set the log image scale factor
__cfg__.BV_LOG_FACTOR = 4

# For the VFE layer
__cfg__.VFE_OUT_DIMS = [32,128]
__cfg__.VFE_FINAl_OUT_DIM = 128

# cal mean from train set
__cfg__.MATRIX_P2 = ([[719.787081,    0.,            608.463003, 44.9538775],
                  [0.,            719.787081,    174.545111, 0.1066855],
                  [0.,            0.,            1.,         3.0106472e-03],
                  [0.,            0.,            0.,         0]])

# cal mean from train set
__cfg__.MATRIX_T_VELO_2_CAM = ([
    [7.49916597e-03, -9.99971248e-01, -8.65110297e-04, -6.71807577e-03],
    [1.18652889e-02, 9.54520517e-04, -9.99910318e-01, -7.33152811e-02],
    [9.99882833e-01, 7.49141178e-03, 1.18719929e-02, -2.78557062e-01],
    [0, 0, 0, 1]
])
# cal mean from train set
__cfg__.MATRIX_R_RECT_0 = ([
    [0.99992475, 0.00975976, -0.00734152, 0],
    [-0.0097913, 0.99994262, -0.00430371, 0],
    [0.00729911, 0.0043753, 0.99996319, 0],
    [0, 0, 0, 1]
])


# Faster-RCNN/SSD Hyper params
if __cfg__.DETECT_OBJECT == 'Car':
    # car anchor
    __cfg__.ANCHOR_L = 3.9
    __cfg__.ANCHOR_W = 1.6
    __cfg__.ANCHOR_H = 1.56
    __cfg__.ANCHOR_Z = -1.0 - __cfg__.ANCHOR_H/2
    __cfg__.RPN_POS_IOU = 0.6
    __cfg__.RPN_NEG_IOU = 0.45

elif __cfg__.DETECT_OBJECT == 'Pedestrian':
    # pedestrian anchor
    __cfg__.ANCHOR_L = 0.8
    __cfg__.ANCHOR_W = 0.6
    __cfg__.ANCHOR_H = 1.73
    __cfg__.ANCHOR_Z = -0.6 - __cfg__.ANCHOR_H/2
    __cfg__.RPN_POS_IOU = 0.5
    __cfg__.RPN_NEG_IOU = 0.35

if __cfg__.DETECT_OBJECT == 'Cyclist':
    # cyclist anchor
    __cfg__.ANCHOR_L = 1.76
    __cfg__.ANCHOR_W = 0.6
    __cfg__.ANCHOR_H = 1.73
    __cfg__.ANCHOR_Z = -0.6 - __cfg__.ANCHOR_H/2
    __cfg__.RPN_POS_IOU = 0.5
    __cfg__.RPN_NEG_IOU = 0.35

# for rpn nms
__cfg__.RPN_NMS_POST_TOPK = 20
__cfg__.RPN_NMS_THRESH = 0.1
__cfg__.RPN_SCORE_THRESH = 0.96


__cfg__.CORNER2CENTER_AVG = True  # average version or max version


cfg = __cfg__