-
Notifications
You must be signed in to change notification settings - Fork 0
/
prepare.py
39 lines (31 loc) · 1.82 KB
/
prepare.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
import numpy as np
from functools import reduce
from config import *
from tools import *
voxel_size = timex * np.array([0.4, 0.4, 0.4], dtype=np.float32) # [-4.2, 4.2]x [-2.2, 6.2]x [-4.2, 4.2] (21,21,21)
grid_size = np.array([21, 21, 21], dtype=np.int32)
lidar_coord = timex * np.array([4.2, 2.2, 4.2], dtype=np.float32)
max_particles_voxel = int(reduce(lambda x, y: x * y, voxel_size)/0.001)
mini_voxel_size = np.array([0.2, 0.2, 0.2], dtype=np.float32) # [-4.2, 4.2]x [-4.2, 4.2]x [-0.25, 8.15] (21,21,21)
mini_grid_size = np.array(timex * np.array([44, 44, 44], dtype=np.int32), dtype=np.int32)
mini_lidar_coord = timex * np.array([4.4, 2.4, 4.4], dtype=np.float32)
mini_max_particles_voxel = int(reduce(lambda x, y: x * y, mini_voxel_size)/0.001)
VOXEL_NUMBER = reduce(lambda x, y: x * y, grid_size) # 21*21*21
MINI_VOXEL_NUMBER = reduce(lambda x, y: x * y, mini_grid_size)
GPU_USE_COUNT = len(GPU_AVAILABLE.split(','))
os.makedirs(save_model_dir, exist_ok=True)
os.makedirs(log_dir, exist_ok=True)
os.makedirs(trans_data_dir, exist_ok=True)
train_csv_list = Find_csv_from_parent_folder(csv_folders, csv_range)
# predict_csv_list = Find_csv(predict_dir)
# predict_csv_list.sort(key=get_fps)
bias_x = mini_grid_size[1] * mini_grid_size[2]
bias_y = mini_grid_size[2]
bias_z = 1
bias_voxel = [0, bias_x, -bias_x, bias_y, -bias_y, bias_z, -bias_z,
bias_x + bias_y, bias_x - bias_y, -bias_x + bias_y, -bias_x - bias_y,
bias_x + bias_z, bias_x - bias_z, -bias_x + bias_z, -bias_x - bias_z,
bias_y + bias_z, bias_y - bias_z, -bias_y + bias_z, -bias_y - bias_z,
bias_x + bias_y + bias_z, bias_x + bias_y - bias_z, bias_x - bias_y + bias_z, bias_x - bias_y - bias_z,
-bias_x + bias_y + bias_z, -bias_x + bias_y - bias_z, -bias_x - bias_y + bias_z,
-bias_x - bias_y - bias_z]