-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathgenerate_shapes.py
181 lines (154 loc) · 8.67 KB
/
generate_shapes.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
# adapted from https://github.com/1zb/3DShape2VecSet/blob/master/sample_class_cond.py
import argparse
import math
import numpy as np
import mcubes
import torch
import trimesh
from pathlib import Path
from models.var_ae import KLAutoEncoder, StandarScaler
from models.ldm import EDMPrecondSkel, EDMPrecondSkelCloud
import os
from models.skelnet import SkelAutoencoder
from models.vecset_encoder import VecSetEncoder
from utils.reconstruction import get_mesh_from_latent_combination
from utils.visual import get_o3d_mesh, get_o3d_cloud
import sys
from time import time
import warnings
def count_parameters(model):
return sum(p.numel() for p in model.parameters() if p.requires_grad)
if __name__ == "__main__":
# category mapper
category_ids = {
'02691156': 0, '02747177': 1, '02773838': 2, '02801938': 3, '02808440': 4, '02818832': 5,
'02828884': 6, '02843684': 7, '02871439': 8, '02876657': 9, '02880940': 10, '02924116': 11,
'02933112': 12, '02942699': 13,'02946921': 14, '02954340': 15, '02958343': 16, '02992529': 17,
'03001627': 18, '03046257': 19, '03085013': 20, '03207941': 21, '03211117': 22, '03261776': 23,
'03325088': 24, '03337140': 25, '03467517': 26, '03513137': 27, '03593526': 28, '03624134': 29,
'03636649': 30, '03642806': 31, '03691459': 32, '03710193': 33, '03759954': 34, '03761084': 35,
'03790512': 36, '03797390': 37, '03928116': 38, '03938244': 39, '03948459': 40, '03991062': 41,
'04004475': 42, '04074963': 43, '04090263': 44, '04099429': 45, '04225987': 46, '04256520': 47,
'04330267': 48, '04379243': 49, '04401088': 50, '04460130': 51, '04468005': 52, '04530566': 53,
'04554684': 54,
}
parser = argparse.ArgumentParser('', add_help=False)
parser.add_argument('--ae-pth', type=str, required=True)
parser.add_argument('--skel_dm_pth', type=str, required=True)
parser.add_argument('--latent_dm_pth', type=str, required=True)# 'output/uncond_dm/kl_d512_m512_l16_edm/checkpoint-999.pth'
#parser.add_argument('--latent_encoder_path', type=str, required=True)
parser.add_argument('--kl_latent_dim', type=int, required=True)
parser.add_argument('--num_skel_samples', type=int, default=512)
parser.add_argument('--num_samples', type=int, default=5)
parser.add_argument('--num_iters', type=int, default=5)
parser.add_argument('--skel_seed', type=int, default=0)
parser.add_argument('--latent_seed', type=int, default=0)
parser.add_argument('--seed_offset', type=int, default=0)
parser.add_argument('--resolution', type=int, default=128)
parser.add_argument('--point_bs', type=int, default=50000)
parser.add_argument('--skel_nn', type=int, default=1)
parser.add_argument('--trunc_val', type=float, default=None)
parser.add_argument('--out_folder', type=str, required=True)
parser.add_argument('--category', type=str, default=None)
parser.add_argument('--check_exist', action='store_true', default=False)
parser.add_argument('--use_spherical_reconstruction', action='store_true', default=False)
parser.add_argument('--ball_nn', type=int, default=2)
parser.add_argument('--ball_margin', type=float, default=0.02)
parser.add_argument('--save_skeleton', action='store_true', default=False)
parser.add_argument('--use_standard_scaler', action='store_true', default=False)
parser.add_argument('--load_skeleton', action='store_true', default=False)
parser.add_argument('--generate_only_skeleton', action='store_true', default=False)
parser.add_argument('--skel_suffix', type=str, default='')
parser.add_argument('--surf_suffix', type=str, default='')
parser.add_argument('--device', type=str, default='cuda:0')
parser.add_argument('--num_skel_steps', type=int, default=20)
parser.add_argument('--num_surf_steps', type=int, default=20)
args = parser.parse_args()
#print(args)
out_file_path = args.out_folder + f'/{args.category}-{args.skel_seed}-{args.latent_seed}{args.surf_suffix}.ply'
if os.path.exists(out_file_path) and args.check_exist:
print('Mesh exists, skipping generation')
sys.exit()
else:
print('Generating shape', out_file_path)
device = args.device
skel_model = EDMPrecondSkelCloud(n_latents=args.num_skel_samples, channels=3, depth=24)
os.makedirs(args.out_folder, exist_ok=True)
skel_model.load_state_dict(torch.load(args.skel_dm_pth)['model'])
skel_model.eval()
skel_model.to(device)
category_id = category_ids[args.category]
skel_model.n_latents = args.num_skel_samples
# seed offset for evaluation purposes
seed_offset = args.seed_offset*10000
# skeleton generation
# load skeleton option is for precomputed skeletons
if args.load_skeleton:
skel_path = args.out_folder + f'/{args.category}_gen_skel_seed_{args.skel_seed}{args.skel_suffix}.npy'
print('Loading sampled skeletons from ', skel_path)
sampled_skeleton = np.load(skel_path)
sampled_skeleton = sampled_skeleton*2
sampled_skeleton = torch.FloatTensor(sampled_skeleton)
print('Skeleton shape:', sampled_skeleton.shape)
else:
with torch.no_grad():
sampled_skeleton = skel_model.sample(None, labels=torch.Tensor([category_id]).long().to(device),
batch_seeds=torch.LongTensor([args.skel_seed + seed_offset]).to(device),
num_steps=args.num_skel_steps,
trunc_val=args.trunc_val).float()
print('Skeleton shape:', sampled_skeleton.shape)
sampled_skeleton = sampled_skeleton/2
if args.save_skeleton and not args.load_skeleton:
skel_path = args.out_folder + f'/{args.category}_gen_skel_seed_{args.skel_seed}{args.skel_suffix}.npy'
print('Saving sampled skeletons to', skel_path)
np.save(skel_path, sampled_skeleton.detach().cpu().numpy())
del skel_model
torch.cuda.empty_cache()
if args.generate_only_skeleton:
sys.exit('Was instructed to generate only skeleton.')
if args.use_standard_scaler:
latent_encoder_object = StandarScaler
else:
latent_encoder_object = KLAutoEncoder
latent_model = EDMPrecondSkel(n_latents=args.num_skel_samples, channels=256, depth=24,
latent_encoder_object=latent_encoder_object,
latent_dim=args.kl_latent_dim)
latent_model.load_state_dict(torch.load(args.latent_dm_pth)['model'])
latent_model.eval()
latent_model.to(device)
sampled_skeleton = sampled_skeleton.to(device)
# surface generation
with torch.no_grad():
sampled_latents = latent_model.sample(skeletons=sampled_skeleton,
labels=torch.Tensor([category_id]).long().to(device),
batch_seeds=torch.LongTensor([args.latent_seed + seed_offset]).to(device),
num_steps=args.num_surf_steps).float()
decoded_latents = latent_model.latent_encoder.decode(sampled_latents)
#saving some memory
del latent_model
torch.cuda.empty_cache()
### reconstruction
ae = SkelAutoencoder(use_skel_model=False, surface_num=2048,
encoder_object=VecSetEncoder,
skel_model_object=None,
num_encoder_heads=8,
num_disps=1)
ae.eval()
ae.load_state_dict(torch.load(args.ae_pth, map_location='cpu'))
ae.to(device)
o3dmesh, recon_mesh, IF = get_mesh_from_latent_combination(decoded_latents, sampled_skeleton,
ae,
args.resolution,
level=0.0, skel_nn=args.skel_nn,
padding=8,
max_dimensions=0.5 * np.array([1, 1, 1]),
min_dimensions=0.5 * np.array([-1, -1, -1]),
bs=args.point_bs,
use_spherical_reconstruction=args.use_spherical_reconstruction,
ball_nn=args.ball_nn,
ball_margin=args.ball_margin)
recon_mesh.vertices *= 2
mesh_path = args.out_folder + f'/{args.category}-{args.skel_seed}-{args.latent_seed}{args.surf_suffix}.ply'
print('Saving final mesh to', mesh_path)
recon_mesh.export(mesh_path)
print('Done')