-
Notifications
You must be signed in to change notification settings - Fork 80
/
make_canonical.py
182 lines (162 loc) · 6.63 KB
/
make_canonical.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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
import numpy as np
import os,sys,copy,glob,cv2,trimesh,time,shutil,pickle,gzip,logging,argparse,difflib
from sklearn.cluster import DBSCAN
logging.getLogger().setLevel(logging.FATAL)
code_dir = os.path.dirname(os.path.realpath(__file__))
sys.path.append(code_dir)
sys.path.append('{}/ss-pybullet'.format(code_dir))
import warnings
warnings.filterwarnings("ignore")
import open3d as o3d
from PIL import Image
from transformations import *
from pybullet_env.utils_pybullet import *
from pybullet_env.env import *
from pybullet_env.env_grasp import *
import pybullet as p
import pybullet_data
from Utils import *
from data_reader import *
import pybullet_tools.utils as PU
import matplotlib.pyplot as plt
from dexnet.grasping.gripper import RobotGripper
from autolab_core import YamlConfig
from dexnet.grasping.gripper import save_grasp_pose_mesh
from renderer import ModelRendererOffscreen
import mpl_toolkits.mplot3d.axes3d as p3
from multiprocessing import Pool
import multiprocessing
from functools import partial
from itertools import repeat
import itertools
try:
multiprocessing.set_start_method('spawn')
except:
pass
def compute_canonical_model():
obj_files = []
names = cfg['dataset'][class_name]['train']
code_dir = os.path.dirname(os.path.realpath(__file__))
for name in names:
obj_files.append(f'{code_dir}/data/object_models/{name}')
out_file = f'{code_dir}/data/object_models/{class_name}_canonical.pkl'
print("obj_files:\n", '\n'.join(obj_files))
clouds = {}
file2normals = {}
for file in obj_files:
mesh = trimesh.load(file)
pts, face_ids = trimesh.sample.sample_surface_even(mesh, 20000)
normals = mesh.face_normals[face_ids]
file2normals[file] = copy.deepcopy(normals)
pcd = toOpen3dCloud(pts)
pcd.voxel_down_sample(voxel_size=0.001)
clouds[file] = np.asarray(pcd.points).copy()
transforms_to_nocs = {}
for obj_file in obj_files:
max_xyz = clouds[obj_file].max(axis=0)
min_xyz = clouds[obj_file].min(axis=0)
transforms_to_nocs[obj_file] = np.eye(4)
nunocs_scale = 1.
center = (max_xyz+min_xyz)/2
transforms_to_nocs[obj_file][:3,3] = -center
transforms_to_nocs[obj_file][:3,:3] = np.diag(np.ones((3))/(max_xyz-min_xyz)) / nunocs_scale
pcd = toOpen3dCloud(clouds[obj_file])
pcd.transform(transforms_to_nocs[obj_file])
max_xyz = np.asarray(pcd.points).max(axis=0)
min_xyz = np.asarray(pcd.points).min(axis=0)
new_tf = np.eye(4)
new_tf[:3,3] = -(max_xyz+min_xyz)/2
transforms_to_nocs[obj_file] = new_tf@transforms_to_nocs[obj_file]
dist_to_other_models = {}
for i,obj_file in enumerate(obj_files):
dists = []
cloud = copy.deepcopy(clouds[obj_file])
cloud = (transforms_to_nocs[obj_file]@to_homo(cloud).T).T[:,:3]
for other_file in obj_files:
if obj_file==other_file:
continue
other_cloud = copy.deepcopy(clouds[other_file])
other_cloud = (transforms_to_nocs[other_file]@to_homo(other_cloud).T).T[:,:3]
cd = chamfer_distance_between_clouds_mutual(cloud,other_cloud)
dists.append(cd)
avg_dist = np.concatenate(dists,axis=0).reshape(-1).mean()
dist_to_other_models[obj_file] = avg_dist
best_dist_id = np.array(dist_to_other_models.values()).argmin()
best_file = list(dist_to_other_models.keys())[best_dist_id]
print('best_file',best_file)
canonical_cloud = copy.deepcopy(clouds[best_file])
canonical_normals = copy.deepcopy(file2normals[best_file])
pcd = toOpen3dCloud(canonical_cloud,normals=canonical_normals)
pcd.transform(transforms_to_nocs[best_file])
canonical_cloud = np.asarray(pcd.points).copy()
canonical_normals = np.asarray(pcd.normals).copy()
############ Gather grasp codebook
gripper = RobotGripper.load(gripper_dir=cfg_grasp['gripper_dir'][class_name])
canonical_grasps = []
grasp_score_thres = 0.8
for obj_file in obj_files:
grasp_file = obj_file.replace('.obj','_complete_grasp.pkl')
with gzip.open(grasp_file,'rb') as ff:
grasps = pickle.load(ff)
new_grasps = []
print(f"Before filter score, #grasps={len(grasps)}")
for grasp in grasps:
if grasp.perturbation_score<grasp_score_thres:
continue
new_grasps.append(grasp)
grasps = new_grasps
grasps = np.array(grasps)
print(f"After filter score, #grasps={len(grasps)}")
for grasp in grasps:
canonical_grasp = copy.deepcopy(grasp)
grasp_pose = grasp.get_grasp_pose_matrix()
canonical_grasp.grasp_pose = transforms_to_nocs[obj_file]@grasp_pose
canonical_grasp.grasp_pose = normalizeRotation(canonical_grasp.grasp_pose)
canonical_grasps.append(canonical_grasp)
canonical_grasps = np.array(canonical_grasps)
############ Gather affordance codebook
print("Gathering affordance")
canonical_affordance = np.zeros((len(canonical_cloud)))
n_afford = 0
for obj_file in obj_files:
print("obj_file:",obj_file)
affordance_file = obj_file.replace('.obj','_affordance.ply')
affordance_pcd = o3d.io.read_point_cloud(affordance_file)
affordance_pcd.transform(transforms_to_nocs[obj_file])
afforance_pts = np.asarray(affordance_pcd.points).copy()
affordance_probs = np.asarray(affordance_pcd.colors)[:,0].copy()
kdtree = cKDTree(afforance_pts)
print("affordance_probs",affordance_probs.min(),affordance_probs.max())
dists,indices = kdtree.query(canonical_cloud)
canonical_affordance += affordance_probs[indices]
n_afford += 1
canonical_affordance /= n_afford
afford_max = canonical_affordance.max()
afford_min = canonical_affordance.min()
print("canonical afford_min: {}, afford_max: {}".format(afford_min,afford_max))
colors = array_to_heatmap_rgb(canonical_affordance.reshape(-1))
pcd = toOpen3dCloud(canonical_cloud,colors)
o3d.io.write_point_cloud(out_file.replace('.pkl','_affordance_vis.ply'),pcd,write_ascii=True)
print(f"Write to {out_file}")
with gzip.open(out_file,'wb') as ff:
out = {
'obj_files': obj_files,
'canonical_cloud': canonical_cloud,
'canonical_normals': canonical_normals,
'transforms_to_nocs': transforms_to_nocs,
'canonical_grasps': canonical_grasps,
'canonical_affordance': canonical_affordance,
}
print(f'#canonical_grasps={len(canonical_grasps)}')
pickle.dump(out,ff)
if __name__=="__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--class_name',type=str,default='nut')
args = parser.parse_args()
code_dir = os.path.dirname(os.path.realpath(__file__))
with open('{}/config.yml'.format(code_dir),'r') as ff:
cfg = yaml.safe_load(ff)
cfg_grasp = YamlConfig("{}/config_grasp.yml".format(code_dir))
class_name = args.class_name
gripper = RobotGripper.load(gripper_dir=cfg_grasp['gripper_dir'][class_name])
compute_canonical_model()