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
183
184
185
186 | import shutil
import os
import sys
import random
import numpy as np
import torch
from torch.utils.data import Dataset
import potpourri3d as pp3d
sys.path.append(os.path.join(os.path.dirname(__file__), "../../src/"))
import fannet
from fannet.utils import toNP
class Shrec11MeshDataset_Original(Dataset):
# NOTE: Original data from the challenge, not simplified models
# The original SHREC11 models were previously distributed via NIST [here](https://www.nist.gov/itl/iad/shrec-2011-datasets), but that page seems to have been lost to the sands of time. We provide a zip of the old dataset page here: https://drive.google.com/uc?export=download&id=1O_P03aAxhjCOKQH2n71j013-EfSmEp5e. The relevant files are in `SHREC11_test_database_new.zip`, which is password protected with the password `SHREC11@NIST`.
# Unzip it like
# unzip -P SHREC11@NIST SHREC11_test_database_new.zip -d [DATA_ROOT]/raw
def __init__(self, root_dir, split_size, k_eig, exclude_dict=None, op_cache_dir=None, spoke_length=[0.00,0.01,0.02,0.03]):
self.root_dir = root_dir
self.n_class = 30
self.split_size = split_size # pass None to take all entries (except those in exclude_dict)
self.k_eig = k_eig
self.op_cache_dir = op_cache_dir
self.spoke_length = spoke_length
self.class_names = []
self.entries = {}
# store in memory
self.verts_list = []
self.faces_list = []
self.labels_list = []
raw_path = os.path.join(self.root_dir, 'raw')
## Parse the categories file
cat_path = os.path.join(self.root_dir, 'categories.txt')
with open(cat_path) as cat_file:
cat_file.readline() # skip the first two lines
cat_file.readline()
for i_class in range(30):
cat_file.readline()
class_name, _, count = cat_file.readline().strip().split()
count = int(count)
self.class_names.append(class_name)
mesh_list = []
for j in range(20):
mesh_list.append(cat_file.readline().strip())
# Randomly grab samples for this split. If given, disallow any samples in commmon with exclude_dict (ie making sure train set is distinct from test).
order = np.random.permutation(len(mesh_list))
added = 0
self.entries[class_name] = set()
for ind in order:
if(split_size is not None and added == split_size): continue
name = mesh_list[ind]
if exclude_dict is not None and name in exclude_dict[class_name]:
continue
path = os.path.join(root_dir, "raw", "T{}.off".format(name))
verts, faces = pp3d.read_mesh(path)
verts = torch.tensor(verts).float()
faces = torch.tensor(faces)
#print(verts)
#print(faces)
# center and unit scale
verts = fannet.geometry.normalize_positions(verts)
self.verts_list.append(verts)
self.faces_list.append(faces)
self.labels_list.append(i_class)
self.entries[class_name].add(name)
added += 1
print(class_name + " -- " + " ".join([p for p in self.entries[class_name]]))
if(split_size is not None and added < split_size):
raise ValueError("could not find enough entries to generate requested split")
for ind, label in enumerate(self.labels_list):
self.labels_list[ind] = torch.tensor(label)
def __len__(self):
return len(self.verts_list)
def __getitem__(self, idx):
verts = self.verts_list[idx]
faces = self.faces_list[idx]
label = self.labels_list[idx]
frames, mass, L, evals, evecs, gradX, gradY, fanIndices = fannet.geometry.get_operators(verts, faces, k_eig=self.k_eig, op_cache_dir=self.op_cache_dir, spoke_length=self.spoke_length)
return verts, faces, frames, mass, L, evals, evecs, gradX, gradY, label , fanIndices
class Shrec11MeshDataset_Simplified(Dataset):
# NOTE: Remeshed data from MeshCNN authors.
# Can be downloaded here (link from the MeshCNN authors): https://www.dropbox.com/s/w16st84r6wc57u7/shrec_16.tar.gz. Note that despite the filename, this really is the shapes from the SHREC 2011 dataset. Extract it to the `[ROOT_DIR]/raw/` directory.
def __init__(self, root_dir, split_size, k_eig, exclude_dict=None, op_cache_dir=None, spoke_length=[0.00,0.01,0.02,0.03]):
self.root_dir = root_dir
self.n_class = 30
self.split_size = split_size # pass None to take all entries (except those in exclude_dict)
self.k_eig = k_eig
self.op_cache_dir = op_cache_dir
self.spoke_length = spoke_length
self.class_names = [ 'alien', 'ants', 'armadillo', 'bird1', 'bird2', 'camel', 'cat', 'centaur', 'dinosaur', 'dino_ske', 'dog1', 'dog2', 'flamingo', 'glasses', 'gorilla', 'hand', 'horse', 'lamp', 'laptop', 'man', 'myScissor', 'octopus', 'pliers', 'rabbit', 'santa', 'shark', 'snake', 'spiders', 'two_balls', 'woman']
self.entries = {}
# store in memory
self.verts_list = []
self.faces_list = []
self.labels_list = []
raw_path = os.path.join(self.root_dir, 'raw', "shrec_16")
for class_idx, class_name in enumerate(self.class_names):
# load both train and test subdirectories; we are manually regenerating random splits to do multiple trials
mesh_files = []
for t in ['test', 'train']:
files = os.listdir(os.path.join(raw_path, class_name, t))
for f in files:
full_f = os.path.join(raw_path, class_name, t, f)
mesh_files.append(full_f)
# Randomly grab samples for this split. If given, disallow any samples in commmon with exclude_dict (ie making sure train set is distinct from test).
order = np.random.permutation(len(mesh_files))
added = 0
self.entries[class_name] = set()
for ind in order:
if(split_size is not None and added == split_size): continue
path = mesh_files[ind]
if exclude_dict is not None and path in exclude_dict[class_name]:
continue
verts, faces = pp3d.read_mesh(path)
verts = torch.tensor(verts).float()
faces = torch.tensor(faces)
#print(verts.shape)
#print(faces.shape)
# center and unit scale
verts = fannet.geometry.normalize_positions(verts)
self.verts_list.append(verts)
self.faces_list.append(faces)
self.labels_list.append(class_idx)
self.entries[class_name].add(path)
added += 1
print(class_name + " -- " + " ".join([os.path.basename(p) for p in self.entries[class_name]]))
if(split_size is not None and added < split_size):
raise ValueError("could not find enough entries to generate requested split")
for ind, label in enumerate(self.labels_list):
self.labels_list[ind] = torch.tensor(label)
# Precompute operators
self.frames_list, self.massvec_list, self.L_list, self.evals_list, self.evecs_list, self.gradX_list, self.gradY_list, self.fanIndices = fannet.geometry.get_all_operators(self.verts_list,
self.faces_list, k_eig=self.k_eig, op_cache_dir=self.op_cache_dir, spoke_length=self.spoke_length)
def __len__(self):
return len(self.verts_list)
def __getitem__(self, idx):
return self.verts_list[idx], self.faces_list[idx], self.frames_list[idx], self.massvec_list[idx], self.L_list[idx], self.evals_list[idx], self.evecs_list[idx], self.gradX_list[idx], self.gradY_list[idx], self.labels_list[idx], self.fanIndices[idx]
|