-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathmain.py
160 lines (130 loc) · 6.99 KB
/
main.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
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
import os
import sys
import json
import torch
import hydra
import trimesh
import numpy as np
import open3d as o3d
import os.path as osp
from tqdm import trange
from typing import Tuple
from omegaconf import DictConfig
sys.path.append('.')
from utils import visualize_point_cloud, load_obj, export_meshes, rfu_filter
from core.config import AlgoConfig
from core.geometry.gripper import Gripper
from core.geometry.samplePosition import sample_position
from core.collision.graspCollisionCheckAvoidCollision import grasp_collision_check_avoid_collision
from core.dipgrasp import dipgrasp
torch.set_default_dtype(torch.float)
def generate_grasp_for_obj(point_cloud: o3d.geometry.PointCloud, gripper: Gripper, cfg: AlgoConfig, sample_num: int = 0) -> Tuple[np.ndarray, np.ndarray]:
if sample_num > 0:
cfg.total_num = sample_num
device = cfg.device
Tbase = torch.eye(4).to(device)
src_points, src_normals, src_weights = gripper.compute_pcd(Tbase.unsqueeze(0),
cfg.joint_init_value.unsqueeze(0),
cfg.link_partial_points,
cfg.link_partial_normals,
True)
dest_points = torch.from_numpy(np.array(point_cloud.points))
dest_normals = torch.from_numpy(np.array(point_cloud.normals))
if dest_points.shape[0] > cfg.algo_params.total_num:
downsample_idx = torch.randperm(dest_points.shape[0])[:cfg.total_num]
dest_points = dest_points[downsample_idx]
dest_normals = dest_normals[downsample_idx]
points = dest_points.numpy()
normals = dest_normals.numpy()
idx = 0
idx_list = [0]
while idx < points.shape[0]:
idx += cfg.algo_params.sample_time
if idx > points.shape[0]:
idx = points.shape[0]
idx_list.append(idx)
pose_list = []
joint_list = []
for i in trange(len(idx_list) - 1):
start_idx, end_idx = idx_list[i], idx_list[i + 1]
repeat_times = end_idx - start_idx
cfg.dest_normals = dest_normals.unsqueeze(0).repeat(repeat_times, 1, 1).to(device).to(cfg.dtype)
cfg.dest_points = dest_points.unsqueeze(0).repeat(repeat_times, 1, 1).to(device).to(cfg.dtype)
sample_points = torch.from_numpy(points[start_idx: end_idx]).to(device).to(cfg.dtype)
sample_normals = torch.from_numpy(normals[start_idx: end_idx]).to(device).to(cfg.dtype)
## Generate initial pose and copy source point cloud
src_points_, src_normals_, sample_T = sample_position(src_points,
src_normals,
sample_points,
sample_normals,
cfg)
src_weights_ = src_weights.repeat((repeat_times, 1))
final_pose, final_joint = dipgrasp(src_points_,
src_normals_,
src_weights_,
sample_T,
gripper,
cfg)
collision_point_num = grasp_collision_check_avoid_collision(final_pose, final_joint, gripper, cfg)
collision_free_idx = collision_point_num < cfg.algo_params.collision_filter_threshold
final_pose = final_pose[collision_free_idx]
final_joint = final_joint[collision_free_idx]
if final_pose.shape[0] == 0: continue
if len(pose_list) == 0:
pose_list = np.asarray(final_pose.detach().cpu().numpy())
joint_list = np.asarray(final_joint.detach().cpu().numpy())
else:
pose_list = np.vstack((pose_list, np.asarray(final_pose.detach().cpu().numpy())))
joint_list = np.vstack((joint_list, np.asarray(final_joint.detach().cpu().numpy())))
if cfg.visualization_params.visualize:
src_points_vis, src_normals_vis, _ = gripper.compute_pcd(final_pose,
final_joint,
cfg.link_complete_points,
cfg.link_complete_normals)
for idx, src_points_vis_ in enumerate(src_points_vis):
visualize_point_cloud(src_points_vis_.detach().cpu().numpy(), dest_points)
return pose_list, joint_list
@hydra.main(version_base="v1.2", config_path='conf', config_name='default')
def main(cfg: DictConfig):
assert osp.exists(cfg.datafile), 'The input file is not exist'
if not osp.exists(cfg.savepath):
os.makedirs(cfg.savepath)
device = cfg.device
data_file = cfg.datafile
save_path = cfg.savepath
gripper_name = cfg.gripper.name
assert gripper_name in ['barrett', 'svh', 'shadow']
gripper = Gripper(gripper_name)
algo_param_dict = DictConfig({**cfg.algo_params, **cfg.gripper.optim_params})
algo_cfg = AlgoConfig(cfg.visualize_setting, algo_param_dict, device=device)
algo_cfg.init_joint_param(gripper)
algo_cfg.init_sample_config(gripper)
algo_cfg.link_partial_points, algo_cfg.link_partial_normals = gripper.get_link_pcd_from_xml()
algo_cfg.link_complete_points, algo_cfg.link_complete_normals = gripper.get_link_pcd_from_mesh()
# sample point
ptCloud = load_obj(data_file)
poses, joints = generate_grasp_for_obj(ptCloud, gripper, algo_cfg)
# convex decomposition
if cfg.simulator:
origin_obj = trimesh.load(data_file)
vhacd_objs = trimesh.decomposition.convex_decomposition(origin_obj, maxConvexHulls=32)
_, filename = os.path.split(data_file)
tmp_path = osp.join(save_path, 'tmp')
os.makedirs(tmp_path, exist_ok=True)
vhacd_path = osp.join(tmp_path, 'vhacd_' + filename)
export_meshes(vhacd_objs, vhacd_path)
poses, joints = rfu_filter(poses, joints, gripper_name, vhacd_path)
if cfg.visualize_setting.visualize_after_simulator:
poses_ = torch.from_numpy(poses).to(device).float()
joints_ = torch.from_numpy(joints).to(device).float()
src_points_vis, src_normals_vis, _ = gripper.compute_pcd(poses_,
joints_,
algo_cfg.link_complete_points,
algo_cfg.link_complete_normals)
dest_points = torch.from_numpy(np.array(ptCloud.points))
for idx, src_points_vis_ in enumerate(src_points_vis):
visualize_point_cloud(src_points_vis_.detach().cpu().numpy(), dest_points)
np.save(osp.join(save_path, 'pose.npy'), poses)
np.save(osp.join(save_path, 'joint_state.npy'), joints)
if __name__ == '__main__':
main()