fuxi-robot commited on
Commit
ec20fc5
·
verified ·
1 Parent(s): 2bb7853

Update scripts/visualize.py

Browse files
Files changed (1) hide show
  1. scripts/visualize.py +107 -37
scripts/visualize.py CHANGED
@@ -4,30 +4,94 @@ import h5py
4
  import imageio
5
  import argparse
6
  import numpy as np
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7
 
8
 
9
- METER_PER_PIXEL_X = 0.2
10
- METER_PER_PIXEL_Y = 0.2
11
- BUCKET_HALF_X = 2
12
- BUCKET_HALF_Y = 2
13
- MAIN_BGR_H = 384
14
- MAIN_BGR_W = 480
15
-
16
-
17
- def excavator_kinematic(qpos, k=6.670, n=2.90746):
18
- '''根据关节角度、大小臂长度计算末端位姿'''
19
- # 490: k=6.670, n=2.90746
20
- boom = qpos[:, 0]
21
- arm = qpos[:, 1]
22
- bucket = qpos[:, 2]
23
- swing = qpos[:, 3]
24
- x = k * np.sin(boom + 1) + n * np.sin(np.pi - boom - 1 - arm - 1.5)
25
- z = k * np.cos(boom + 1) - n * np.cos(np.pi - boom - 1 - arm - 1.5)
26
-
27
- return x, z, swing, bucket
28
-
29
-
30
- def resize_with_aspect_ratio(image, width=None, height=None):
31
  if width is None and height is None:
32
  return image
33
 
@@ -51,36 +115,43 @@ def resize_with_aspect_ratio(image, width=None, height=None):
51
  return padded
52
 
53
 
54
- def traj_visualization(data_path, out_dir, fmt='mp4', fps=30):
 
 
 
 
 
 
55
  # load data
56
  with h5py.File(data_path) as f:
57
  elevations = np.array(f['observations']['images']['elevation'])
58
  joints = np.array(f['observations']['qpos'])
59
  mains = np.array(f['observations']['images']['main'])
60
-
61
  assert elevations.shape[0] == joints.shape[0]
62
- H, W = elevations.shape[1:3]
63
 
64
  # forward kinemetics
65
- x = excavator_kinematic(joints, k=6.670, n=2.90746)[0]
66
- # x = excavator_kinematic(joints, k=3.6957, n=1.62233)[0]
67
- px = (x / METER_PER_PIXEL_X).astype(np.int32)
 
 
 
68
 
69
  # visualization
70
  frames = []
71
  for i in range(elevations.shape[0]):
72
  # end pose
73
  traj_elevation = elevations[i].copy()
74
- center_u = W // 2
75
- center_v = H // 2 - px[i]
76
- traj_elevation[center_v - BUCKET_HALF_X: center_v + BUCKET_HALF_X,
77
- center_u - BUCKET_HALF_Y: center_u + BUCKET_HALF_Y] = (255, 0, 0)
78
 
79
  # elevation
80
  traj_elevation = resize_with_aspect_ratio(traj_elevation, height=MAIN_BGR_H, width=MAIN_BGR_W)
81
 
82
  # main bgr
83
- traj_main = mains[i].copy()[..., ::-1]
84
  traj_main = resize_with_aspect_ratio(traj_main, height=MAIN_BGR_H, width=MAIN_BGR_W)
85
  traj_img = np.concatenate([traj_main, traj_elevation], axis=1)
86
 
@@ -94,12 +165,11 @@ def traj_visualization(data_path, out_dir, fmt='mp4', fps=30):
94
 
95
  if __name__=="__main__":
96
  parser = argparse.ArgumentParser()
 
97
  parser.add_argument("--data_path", type=str, required=True, help='path to the h5 file')
98
  parser.add_argument("--out_dir", type=str, required=True, help='path the output directory')
99
  parser.add_argument("--format", type=str, required=False, default='mp4', help='format of the output file, mp4 or gif are supported')
100
- parser.add_argument("--fps", type=str, required=False, default=30, help='FPS of the output file')
101
  args = parser.parse_args()
102
 
103
-
104
- "/home/netease/data/excavator_motion/490/robot_data_20250430_123028.h5"
105
- traj_visualization(args.data_path, args.out_dir, args.format)
 
4
  import imageio
5
  import argparse
6
  import numpy as np
7
+ from dataclasses import dataclass
8
+
9
+
10
+ MAIN_BGR_H = 384
11
+ MAIN_BGR_W = 480
12
+ ELEVATION_H = 200
13
+ ELEVATION_W = 200
14
+
15
+
16
+ @dataclass
17
+ class ExcavatorModel:
18
+ name: str = ''
19
+
20
+ # body
21
+ boom_len: float = 0.0
22
+ arm_len: float = 0.0
23
+ operating_arm_height: float = 0.0
24
+ boom_init_angle: float = 0.0
25
+ arm_init_angle: float = 0.0
26
+ arm_offset_x: float = 0.0
27
+ arm_offset_y: float = 0.0
28
+
29
+ # bucket
30
+ bucket_length: float = 0.0
31
+ bucket_width: float = 0.0
32
+
33
+ # elevation resolution
34
+ meter_per_pixel_u: float = 0.0
35
+ meter_per_pixel_v: float = 0.0
36
+
37
+
38
+ class ExcavatorKinematics(object):
39
+ def __init__(self, excavator: str):
40
+ self.model = self._get_model(excavator)
41
+
42
+ def _get_model(self, excavator: str) -> ExcavatorModel:
43
+ exc = ExcavatorModel(name=excavator)
44
+ if excavator == '75':
45
+ exc.boom_len = 3.6957
46
+ exc.arm_len = 1.62233
47
+ exc.operating_arm_height = 1.4
48
+ exc.boom_init_angle = 1.0
49
+ exc.arm_init_angle = 1.5
50
+ exc.arm_offset_x = 0.0
51
+ exc.arm_offset_y = -0.1
52
+ exc.bucket_length = 0.8
53
+ exc.bucket_width = 0.6
54
+ exc.meter_per_pixel_u = 0.1
55
+ exc.meter_per_pixel_v = 0.1
56
+
57
+ elif excavator in ('490', '306'):
58
+ exc.boom_len = 6.670
59
+ exc.arm_len = 2.90746
60
+ exc.operating_arm_height = 2.46
61
+ exc.boom_init_angle = 1.0
62
+ exc.arm_init_angle = 1.5
63
+ exc.arm_offset_x = 0.0
64
+ exc.arm_offset_y = 0.0
65
+ exc.bucket_length = 1.5
66
+ exc.bucket_width = 1.2
67
+ exc.meter_per_pixel_u = 0.2
68
+ exc.meter_per_pixel_v = 0.2
69
+
70
+ else:
71
+ raise NotImplementedError
72
+
73
+ return exc
74
+
75
+ def fk(self, joints: np.ndarray) -> np.ndarray:
76
+ assert joints.shape[1] == 4
77
+
78
+ boom = joints[:, 0]
79
+ arm = joints[:, 1]
80
+ x = self.model.boom_len * np.sin(boom + self.model.boom_init_angle) + \
81
+ self.model.arm_len * np.sin(np.pi - boom - self.model.boom_init_angle - arm - self.model.arm_init_angle)
82
+ z = self.model.boom_len * np.cos(boom + self.model.boom_init_angle) - \
83
+ self.model.arm_len * np.cos(np.pi - boom - self.model.boom_init_angle - arm - self.model.arm_init_angle)
84
+ y = np.zeros_like(x)
85
+ xyz = np.stack([x, y, z], axis=1)
86
+
87
+ return xyz
88
 
89
 
90
+ def resize_with_aspect_ratio(
91
+ image: np.ndarray,
92
+ width: int=None,
93
+ height: int=None
94
+ ) -> np.ndarray:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
95
  if width is None and height is None:
96
  return image
97
 
 
115
  return padded
116
 
117
 
118
+ def traj_visualization(
119
+ excavator: str,
120
+ data_path: str,
121
+ out_dir: str,
122
+ fmt: str='mp4',
123
+ fps: int=30
124
+ ) -> None:
125
  # load data
126
  with h5py.File(data_path) as f:
127
  elevations = np.array(f['observations']['images']['elevation'])
128
  joints = np.array(f['observations']['qpos'])
129
  mains = np.array(f['observations']['images']['main'])
 
130
  assert elevations.shape[0] == joints.shape[0]
 
131
 
132
  # forward kinemetics
133
+ exc = ExcavatorKinematics(excavator)
134
+ xyz = exc.fk(joints)
135
+ px = (xyz[:, 0] / exc.model.meter_per_pixel_u).astype(np.int32)
136
+
137
+ bucket_half_u = int(exc.model.bucket_length / 2 / exc.model.meter_per_pixel_u)
138
+ bucket_half_v = int(exc.model.bucket_width / 2 / exc.model.meter_per_pixel_v)
139
 
140
  # visualization
141
  frames = []
142
  for i in range(elevations.shape[0]):
143
  # end pose
144
  traj_elevation = elevations[i].copy()
145
+ center_u = ELEVATION_W // 2
146
+ center_v = ELEVATION_H // 2 - px[i]
147
+ traj_elevation[center_v - bucket_half_v: center_v + bucket_half_v,
148
+ center_u - bucket_half_u: center_u + bucket_half_u] = (255, 0, 0)
149
 
150
  # elevation
151
  traj_elevation = resize_with_aspect_ratio(traj_elevation, height=MAIN_BGR_H, width=MAIN_BGR_W)
152
 
153
  # main bgr
154
+ traj_main = mains[i].copy()[..., ::-1] # BGR -> RGB
155
  traj_main = resize_with_aspect_ratio(traj_main, height=MAIN_BGR_H, width=MAIN_BGR_W)
156
  traj_img = np.concatenate([traj_main, traj_elevation], axis=1)
157
 
 
165
 
166
  if __name__=="__main__":
167
  parser = argparse.ArgumentParser()
168
+ parser.add_argument("--excavator", type=str, required=True, help='excavator name')
169
  parser.add_argument("--data_path", type=str, required=True, help='path to the h5 file')
170
  parser.add_argument("--out_dir", type=str, required=True, help='path the output directory')
171
  parser.add_argument("--format", type=str, required=False, default='mp4', help='format of the output file, mp4 or gif are supported')
172
+ parser.add_argument("--fps", type=int, required=False, default=30, help='FPS of the output file')
173
  args = parser.parse_args()
174
 
175
+ traj_visualization(args.excavator, args.data_path, args.out_dir, args.format, args.fps)