-
Notifications
You must be signed in to change notification settings - Fork 1
/
run_nerf_helpers.py
250 lines (207 loc) · 9.49 KB
/
run_nerf_helpers.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
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
import jittor as jt
from jittor import nn
import numpy as np
# Misc
img2mse = lambda x, y : jt.mean((x - y) ** 2)
mse2psnr = lambda x : -10. * jt.log(x) / jt.log(jt.array(np.array([10.])))
to8b = lambda x : (255*np.clip(x,0,1)).astype(np.uint8)
class Embedder:
def __init__(self, **kwargs):
self.kwargs = kwargs
self.create_embedding_fn()
def create_embedding_fn(self):
embed_fns = []
d = self.kwargs['input_dims']
out_dim = 0
if self.kwargs['include_input']:
embed_fns.append(lambda x : x)
out_dim += d
max_freq = self.kwargs['max_freq_log2']
min_freq = self.kwargs['min_freq_log2']
N_freqs = self.kwargs['num_freqs']
if self.kwargs['log_sampling']:
freq_bands = 2.**jt.linspace(min_freq, max_freq, steps=N_freqs)
else:
freq_bands = jt.linspace(2.**min_freq, 2.**max_freq, steps=N_freqs)
for freq in freq_bands:
for p_fn in self.kwargs['periodic_fns']:
embed_fns.append(lambda x, p_fn=p_fn, freq=freq : p_fn(x * freq))
out_dim += d
self.embed_fns = embed_fns
self.out_dim = out_dim
def embed(self, inputs):
return jt.concat([fn(inputs) for fn in self.embed_fns], -1)
class MipEmbedder:
def __init__(self, **kwargs):
self.kwargs = kwargs
self.create_embedding_fn()
def create_embedding_fn(self):
embed_fns = []
d = self.kwargs['input_dims']
out_dim = 0
if self.kwargs['include_input']:
embed_fns.append(lambda x : x[:,:d])
out_dim += d
max_freq = self.kwargs['max_freq_log2']
min_freq = self.kwargs['min_freq_log2']
N_freqs = self.kwargs['num_freqs']
if self.kwargs['log_sampling']:
freq_bands_y = 2.**jt.linspace(min_freq, max_freq, steps=N_freqs)
freq_bands_w = 4.**jt.linspace(min_freq, max_freq, steps=N_freqs)
else:
freq_bands_y = jt.linspace(2.**min_freq, 2.**max_freq, steps=N_freqs)
freq_bands_w = jt.linspace(4.**min_freq, 4.**max_freq, steps=N_freqs)
for ctr in range(len(freq_bands_y)):
for p_fn in self.kwargs['periodic_fns']:
embed_fns.append(lambda inputs, p_fn=p_fn, freq_y=freq_bands_y[ctr], freq_w=freq_bands_w[ctr] : p_fn(inputs[:,:d] * freq_y) * jt.exp((-0.5) * freq_w * inputs[:,d:]))
out_dim += d
self.embed_fns = embed_fns
self.out_dim = out_dim
def embed(self, inputs):
return jt.concat([fn(inputs) for fn in self.embed_fns], -1)
def get_embedder(multires, min_multires=0, i=0, input_dims=3):
if i == -1:
return nn.Identity(), 3
embed_kwargs = {
'include_input' : True,
'input_dims' : input_dims,
'min_freq_log2': min_multires,
'max_freq_log2' : multires-1,
'num_freqs' : multires,
'log_sampling' : True,
'periodic_fns' : [jt.sin, jt.cos]
}
embedder_obj = Embedder(**embed_kwargs)
embed = lambda x, eo=embedder_obj : eo.embed(x)
return embed, embedder_obj.out_dim
def get_mip_embedder(multires, min_multires=0, i=0, input_dims=3):
if i == -1:
return nn.Identity(), 3
embed_kwargs = {
'include_input' : True,
'input_dims' : 3,
'min_freq_log2': min_multires,
'max_freq_log2' : multires-1,
'num_freqs' : multires,
'log_sampling' : True,
'periodic_fns' : [jt.sin, jt.cos],
}
embedder_obj = MipEmbedder(**embed_kwargs)
embed = lambda inputs, eo=embedder_obj : eo.embed(inputs)
return embed, embedder_obj.out_dim
class Bungee_NeRF_baseblock(nn.Module):
def __init__(self, net_width=256, input_ch=3, input_ch_views=3):
super(Bungee_NeRF_baseblock, self).__init__()
self.pts_linears = nn.ModuleList([nn.Linear(input_ch, net_width)] + [nn.Linear(net_width, net_width) for _ in range(3)])
self.views_linear = nn.Linear(input_ch_views + net_width, net_width//2)
self.feature_linear = nn.Linear(net_width, net_width)
self.alpha_linear = nn.Linear(net_width, 1)
self.rgb_linear = nn.Linear(net_width//2, 3)
def execute(self, input_pts, input_views):
h = input_pts
for i, l in enumerate(self.pts_linears):
h = self.pts_linears[i](h)
h = jt.nn.relu(h)
alpha = self.alpha_linear(h)
feature0 = self.feature_linear(h)
h0 = jt.concat([feature0, input_views], -1)
h0 = self.views_linear(h0)
h0 = jt.nn.relu(h0)
rgb = self.rgb_linear(h0)
return rgb, alpha, h
class Bungee_NeRF_resblock(nn.Module):
def __init__(self, net_width=256, input_ch=3, input_ch_views=3):
super(Bungee_NeRF_resblock, self).__init__()
self.pts_linears = nn.ModuleList([nn.Linear(input_ch+net_width, net_width), nn.Linear(net_width, net_width)])
self.views_linear = nn.Linear(input_ch_views + net_width, net_width//2)
self.feature_linear = nn.Linear(net_width, net_width)
self.alpha_linear = nn.Linear(net_width, 1)
self.rgb_linear = nn.Linear(net_width//2, 3)
def execute(self, input_pts, input_views, h):
h = jt.concat([input_pts, h], -1)
for i, l in enumerate(self.pts_linears):
h = self.pts_linears[i](h)
h = jt.nn.relu(h)
alpha = self.alpha_linear(h)
feature0 = self.feature_linear(h)
h0 = jt.concat([feature0, input_views], -1)
h0 = self.views_linear(h0)
h0 = jt.nn.relu(h0)
rgb = self.rgb_linear(h0)
return rgb, alpha, h
class Bungee_NeRF_block(nn.Module):
def __init__(self, num_resblocks=3, net_width=256, input_ch=3, input_ch_views=3):
super(Bungee_NeRF_block, self).__init__()
self.input_ch = input_ch
self.input_ch_views = input_ch_views
self.num_resblocks = num_resblocks
self.baseblock = Bungee_NeRF_baseblock(net_width=net_width, input_ch=input_ch, input_ch_views=input_ch_views)
self.resblocks = nn.ModuleList([
Bungee_NeRF_resblock(net_width=net_width, input_ch=input_ch, input_ch_views=input_ch_views) for _ in range(num_resblocks)
])
def execute(self, x):
input_pts, input_views = jt.split(x, [self.input_ch, self.input_ch_views], dim=-1)
alphas = []
rgbs = []
base_rgb, base_alpha, h = self.baseblock(input_pts, input_views)
alphas.append(base_alpha)
rgbs.append(base_rgb)
for i in range(self.num_resblocks):
res_rgb, res_alpha, h = self.resblocks[i](input_pts, input_views, h)
alphas.append(res_alpha)
rgbs.append(res_rgb)
output = jt.concat([jt.stack(rgbs,1), jt.stack(alphas,1)],-1)
return output
def get_rays(H, W, focal, c2w):
i, j = jt.meshgrid(jt.linspace(0, W-1, W), jt.linspace(0, H-1, H))
i = i.t()
j = j.t()
dirs = jt.stack([(i-W*.5)/focal, -(j-H*.5)/focal, -jt.ones_like(i)], -1)
dirs = dirs/jt.norm(dirs, p=2, dim=-1)[...,None]
rays_d = jt.sum(dirs[..., np.newaxis, :] * c2w[:3,:3], -1)
rays_o = c2w[:3,-1].expand(rays_d.shape)
return rays_o, rays_d
def get_rays_np(H, W, focal, c2w):
i, j = np.meshgrid(np.arange(W, dtype=np.float32), np.arange(H, dtype=np.float32), indexing='xy')
dirs = np.stack([(i-W*.5)/focal, -(j-H*.5)/focal, -np.ones_like(i)], -1)
# Rotate ray directions from camera frame to the world frame
rays_d = np.sum(dirs[..., np.newaxis, :] * c2w[:3,:3], -1) # dot product, equals to: [c2w.dot(dir) for dir in dirs]
# Translate camera frame's origin to the world frame. It is the origin of all rays.
rays_o = np.broadcast_to(c2w[:3,-1], np.shape(rays_d))
return rays_o, rays_d
def get_radii_for_test(H, W, focal, c2w):
i, j = jt.meshgrid(jt.linspace(0, W-1, W), jt.linspace(0, H-1, H))
i = i.t()
j = j.t()
dirs = jt.stack([(i-W*.5)/focal, -(j-H*.5)/focal, -jt.ones_like(i)], -1)
rays_d = jt.sum(dirs[np.newaxis, ..., np.newaxis, :] * c2w[:, np.newaxis, np.newaxis, :3,:3], -1)
dx = jt.sqrt(
jt.sum((rays_d[:, :-1, :, :] - rays_d[:, 1:, :, :])**2, -1))
dx = jt.concat([dx, dx[:, -2:-1, :]], 1)
radii = dx[..., None] * 2 / np.sqrt(12)
return radii
def sample_pdf(bins, weights, N_samples, det=False):
# Get pdf
weights = weights + 1e-5 # prevent nans
pdf = weights / jt.sum(weights, -1, keepdims=True)
cdf = jt.cumsum(pdf, -1)
cdf = jt.concat([jt.zeros_like(cdf[...,:1]), cdf], -1) # (batch, len(bins))
# Take uniform samples
if det:
u = jt.linspace(0., 1., steps=N_samples)
u = u.expand(list(cdf.shape[:-1]) + [N_samples])
else:
u = jt.random(list(cdf.shape[:-1]) + [N_samples])
# Invert CDF
inds = jt.searchsorted(cdf, u, right=True)
below = jt.maximum(jt.zeros_like(inds-1), inds-1)
above = jt.minimum((cdf.shape[-1]-1) * jt.ones_like(inds), inds)
inds_g = jt.stack([below, above], -1) # (batch, N_samples, 2)
matched_shape = [inds_g.shape[0], inds_g.shape[1], cdf.shape[-1]]
cdf_g = jt.gather(cdf.unsqueeze(1).expand(matched_shape), 2, inds_g)
bins_g = jt.gather(bins.unsqueeze(1).expand(matched_shape), 2, inds_g)
denom = (cdf_g[...,1]-cdf_g[...,0])
denom[denom<1e-5] = 1.0
t = (u-cdf_g[...,0])/denom
samples = bins_g[...,0] + t * (bins_g[...,1]-bins_g[...,0])
return samples