-
Notifications
You must be signed in to change notification settings - Fork 3
/
project_texture.py
133 lines (114 loc) · 5.35 KB
/
project_texture.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
import numpy as np
import torch
from PIL import Image
from models.imgsr.utils.common_utils import get_image
from models.layers.mesh_prepare import fill_from_file
from models.losses.loss import ChamferDist
from util.sampler import sampler_uv
import argparse
parser = argparse.ArgumentParser(description='geo proj script',
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('name', metavar='NAME',
help='names')
torch.random.manual_seed(0)
def torch_to_np(img_var):
'''Converts an image in torch.Tensor format to np.array.
From 1 x C x W x H [0..1] to C x W x H [0..1]
'''
return img_var.detach().cpu().numpy()[0]
def get_texture(texture_path):
texture = get_image(texture_path)[1]
return torch.from_numpy(texture).unsqueeze(0).cuda()
def load_obj(path):
obj = {}
vs, faces, vc, uvs, face_uvs, texture, text,_ = fill_from_file(file=path)
obj['vs'] = torch.from_numpy(vs).unsqueeze(0).cuda()
obj['faces'] = torch.from_numpy(faces).unsqueeze(0).cuda()
if uvs is not None:
obj['uvs'] = torch.from_numpy(uvs).unsqueeze(0).cuda()
obj['face_uvs'] = torch.from_numpy(face_uvs).unsqueeze(0).cuda()
if texture is not None:
obj['texture'] = torch.from_numpy(texture).unsqueeze(0).cuda()
if vc is not None:
obj['vc'] = torch.from_numpy(vc).unsqueeze(0).cuda()
else:
obj['vc'] = torch.ones_like(obj['vs'])
return obj
n = 1000000
args = parser.parse_args()
ids = args.name.split(',')
for i in ids:
gt_path = "%s.obj"%i
gt = load_obj(gt_path)
#gt_points, _, gt_uvs = sampler_uv(gt['faces'], gt['vs'], 25000, uvs=gt['uvs'], face_uvs=gt['face_uvs'])
#gt_colors = uv2color(gt_uvs[0], gt['texture'][0])
#save_obj(vs=gt_points[0],faces=[],colors=gt_colors,dir='',filename='pc.obj')
gt_points = gt['vs']
mid_point = torch.mean(gt_points, dim=1, keepdim=True)
gt_dist = torch.sqrt(torch.sum((gt_points-mid_point)**2, dim=-1, keepdim=True))
gt_colors = gt['vc']
mesh_path = "%s_f_uv.obj"%i#"datasets/train/dogt1_ep1_step3000.obj"
#17.90 17.89 17.61
mesh = load_obj(mesh_path)
#mesh['texture'] = get_texture("datasets/result/2/2l_ep1_step1500.jpg")
mesh_points, _, mesh_uvs = sampler_uv(mesh['faces'], mesh['vs'], n, uvs=mesh['uvs'], face_uvs=mesh['face_uvs'])
dist1, dist2, idx1, idx2 = ChamferDist()(gt_points.float(), mesh_points)
pc_uv = mesh_uvs[0][idx1[0].long()].cpu().numpy()
pc_colors = gt_colors[0].cpu().numpy()
pc_xyz = gt_points[0].cpu().numpy()
pc_dist = gt_dist[0].cpu().numpy()
pc_xyz_n = (pc_xyz-np.min(pc_xyz, axis=0, keepdims=True))/(np.max(pc_xyz,
axis=0, keepdims=True)-np.min(pc_xyz, axis=0, keepdims=True)+1e-10)
pc_dist_n = pc_dist#(pc_dist-np.min(pc_dist, axis=0, keepdims=True))/(np.max(pc_dist,
# axis=0, keepdims=True)-np.min(pc_dist, axis=0, keepdims=True)+1e-10)
img_np = np.ones((1024 * 1024, 3))
img_np_xyz = np.ones((1024 * 1024, 3))
img_np_dist = np.zeros((1024 * 1024, 1))
mask = np.zeros((1024 * 1024, 1))
index = np.ones_like(pc_uv[:, 1])
index[:] = np.floor((1-pc_uv[:,1])*1024).astype(np.uint32) * 1024 + np.floor(pc_uv[:,0] * 1024).astype(np.uint32)
index = index.astype(np.long)
img_np[index] = pc_colors
img_np_xyz[index] = pc_xyz_n
img_np_dist[index] = pc_dist_n
mask[index] = 1
# index[:] = (np.clip((1-pc_uv[:,1])*1024+1,0,1023)).astype(np.uint32) * 1024 + (pc_uv[:,0] * 1024).astype(np.uint32)
# #(np.mean(img_np, axis=1)>0).astype(np.float32)
# index = index.astype(np.long)
# img_np[index] = pc_colors
# img_np_xyz[index] = pc_xyz_n
# mask[index] = 1
# index[:] = ((1-pc_uv[:,1])*1024).astype(np.uint32) * 1024 + np.clip((pc_uv[:,0] * 1024+1), 0, 1023).astype(np.uint32)
#
# index = index.astype(np.long)
# img_np[index] = pc_colors
# img_np_xyz[index] = pc_xyz_n
# mask[index] = 1
# index[:] = (np.clip((1-pc_uv[:,1])*1024+1,0,1023)).astype(np.uint32) * 1024 + \
# np.clip((pc_uv[:,0] * 1024+1), 0, 1023).astype(np.uint32)
#
# index = index.astype(np.long)
# img_np[index] = pc_colors
# img_np_xyz[index] = pc_xyz_n
# mask[index] = 1
img_np.resize((1024, 1024, 3))
img_np_xyz.resize((1024, 1024, 3))
img_np_dist.resize((1024, 1024))
ar = np.clip(img_np*255,0,255).astype(np.uint8)
ar_xyz = np.clip(img_np_xyz*255,0,255).astype(np.uint8)
ar_dist = np.clip(img_np_dist*255,0,255).astype(np.uint8)
mask.resize((1024, 1024))
mask_ar = np.clip(mask*255,0,255).astype(np.uint8)
Image.fromarray(ar, mode="RGB").save("input_f/texture%s.jpg"%i)
Image.fromarray(ar_xyz, mode="RGB").save("input_f/xyz%s.jpg"%i)
Image.fromarray(ar_dist).save("input_f/dist%s.jpg"%i)
Image.fromarray(mask_ar).save("input_f/mask%s.jpg"%i)
np.save("input_f/mask%s.npy"%i, mask.astype(np.uint8))
np.save("input_f/texture%s.npy"%i, img_np)
np.save("input_f/xyz%s.npy"%i, img_np_xyz)
np.save("input_f/dist%s.npy"%i, img_np_dist)
print("done")
#mesh_vc = uv2color(obj1_uvs[0], mesh['texture'][0]).unsqueeze(0)
# dist1, dist2, idx1, idx2 = ChamferDist()(obj1_points, mesh_points)
# chamfer1 = torch.mean(torch.abs(mesh_vc[0][idx1[0].long()]-obj1_vc))
# chamfer2 = torch.mean(torch.abs(obj1_vc[0][idx2[0].long()]-mesh_vc))