-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrender.py
111 lines (85 loc) · 3.87 KB
/
render.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
import tensorflow as tf
import numpy as np
def CTRender(M, eview_vec, elight_vec):
""" 对输入的SVBRDF分图，使用微表面模型进行渲染
Args:
M (tensor): 输入SVBRDF四个图片
eview_vec (tensor): 像素到摄像机的分布向量
elight_vec (tensor):像素到光源的分布向量
Returns:
ctrender_batch (tensor): 渲染出的成品图
"""
half_vec = (eview_vec + elight_vec) / 2
half_vec_norm = tf.sqrt(tf.reduce_sum(tf.square(half_vec),axis=-1))
half_vec_expand = tf.expand_dims(half_vec_norm,axis=-1)
newhalf_vec = tf.concat([half_vec_expand,half_vec_expand,half_vec_expand],axis=-1)
ehalf_vec = half_vec/newhalf_vec
norm, diff, rough, spec = tf.split(M, 4, axis=3)
norm = norm*2-1
norm_mo = tf.sqrt(tf.reduce_sum(tf.square(norm),axis=-1))
norm_expand = tf.expand_dims(norm_mo,axis=-1)
newnorm = tf.concat([norm_expand,norm_expand,norm_expand],axis=-1)
enorm = norm/newnorm
NdotH = tf.reduce_sum(tf.multiply(enorm,ehalf_vec),axis=-1) # (None, 256, 256)
nh_expand = tf.expand_dims(NdotH,axis=-1)
nh = tf.concat([nh_expand,nh_expand,nh_expand],axis=-1)
NdotL = tf.reduce_sum(tf.multiply(enorm,elight_vec),axis=-1)
nl_expand = tf.expand_dims(NdotL,axis=-1)
nl = tf.concat([nl_expand,nl_expand,nl_expand],axis=-1)
NdotV = tf.reduce_sum(tf.multiply(enorm,eview_vec),axis=-1)
nv_expand = tf.expand_dims(NdotV,axis=-1)
nv = tf.concat([nv_expand,nv_expand,nv_expand],axis=-1)
VdotH = tf.reduce_sum(tf.multiply(eview_vec,ehalf_vec),axis=-1)
vh_expand = tf.expand_dims(VdotH,axis=-1)
vh = tf.concat([vh_expand,vh_expand,vh_expand],axis=-1)
nh = tf.maximum(nh, 1e-8)
nv = tf.maximum(nv, 1e-8)
nl = tf.maximum(nl, 1e-8)
vh = tf.maximum(vh, 1e-8)
pi = tf.constant(3.1415926)
r2 = rough*rough
denominator = tf.square(nh)*(tf.pow(r2,2)-1)+1+0.0001
Norm_distrib = r2/pi/ denominator
norm_distrib = tf.square(Norm_distrib)*pi
k = tf.maximum(1e-8, rough * rough * 0.5)
shade_mask1 = nl*(1-k)+k
shade_mask2 = nv*(1-k)+k
shade_mask = tf.compat.v1.reciprocal(shade_mask1*shade_mask2)
F_mi = (-5.55473*vh-6.98316)*vh
F_mi = tf.cast(F_mi,tf.float32)
fresnel = spec+(1-spec)*tf.pow(2.0,F_mi)
diff_scale = diff/pi
fr = fresnel*shade_mask*norm_distrib/4 + diff_scale # 微表面模型
ctrender_batch = fr*nl*3.14
return ctrender_batch
def generate_vl(w=3264, h=2448):
""" 生成点到中心光源的向量和光强图
（默认光源位置在图片中心）
Args:
w (int, optional): 图片宽. Defaults to 3264.
h (int, optional): 图片高. Defaults to 2448.
Returns:
eview_vec (tensor): 像素到光源的向量 (1, w, h, 3)
intensity (tensor): intensity分布 (1, w, h, 3)
"""
# d = w/(2*tf.tan(alpha)), alpha = 33/180*pi
d = w/1.29876 # 高度
view_pos = tf.constant([w/2, h/2, d], dtype=tf.float32, shape=[1,3]) # 修改这里
view_pos = tf.expand_dims(view_pos,axis=1)
wgrid = tf.linspace(0.0, w, w)
hgrid = tf.linspace(0.0, h, h)
plane_coor = tf.concat(tf.meshgrid(wgrid,hgrid,0.0),axis=2)
planc_expand = tf.expand_dims(plane_coor,axis=0) #[1, h, w, 3]
# view_vec
view_pos_expand = tf.expand_dims(view_pos,axis=1)
view_vec = view_pos_expand - planc_expand # 图像上每个点到中心光源的向量
view_norm = tf.sqrt(tf.reduce_sum(tf.square(view_vec),axis=-1))
view_expand = tf.expand_dims(view_norm,axis=-1)
newview_vec = tf.concat([view_expand,view_expand,view_expand],axis=-1)
eview_vec = view_vec/newview_vec
# I
xy = planc_expand[:,:,:,0:2]
t = tf.sqrt(tf.reduce_sum(tf.square(xy - tf.constant([w/2, h/2], dtype=tf.float32, shape=[1,2])), keepdims=True, axis=-1))
I = tf.exp(tf.square(tf.tan(t/d*1.6)) * (-0.5))
I = tf.concat([I,I,I], axis=-1)
return eview_vec, I