-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
358 lines (313 loc) · 16.5 KB
/
index.html
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
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>NeILF++: Inter-reflectable Light Fields for Geometry and Material Estimation</title>
<!-- Bootstrap -->
<link href="css/bootstrap-4.4.1.css" rel="stylesheet">
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/4.7.0/css/font-awesome.min.css">
<link rel="stylesheet" href="css/dics.min.css">
<!-- <link rel="stylesheet" href="css/bulma.min.css"> -->
<script src="scripts/dics.min.js"></script>
<script>
document.addEventListener('DOMContentLoaded', domReady);
function domReady() {
for (const e of document.querySelectorAll(".b-dics")) {
new Dics({
container: e,
textPosition: "top"
});
}
}
</script>
</head>
<body>
<section>
<div class="jumbotron text-center mt-4">
<div class="container">
<div class="row">
<div class="col-12">
<h1>NeILF++: Inter-reflectable Light Fields for Geometry and Material Estimation</h1>
<h4 style="color:#5a6268;"> ICCV 2023 </h4>
<hr>
<h5> <a href="https://scholar.google.com/citations?user=diIjfmMAAAAJ&hl=en" target="_blank">Jingyang Zhang<sup>1</sup></a>,
<a href="https://yoyo000.github.io/" target="_blank">Yao Yao<sup>2 <i class="fa fa-envelope"></i></sup></a>,
<a href="https://scholar.google.com.hk/citations?user=YR1MdT0AAAAJ&hl=en" target="_blank">Shiwei Li<sup>1</sup></a>,
<a href="https://scholar.google.com.hk/citations?user=kPZVXzoAAAAJ&hl=en" target="_blank">Jingbo Liu<sup>1</sup></a>,
<a href="https://scholar.google.com.hk/citations?user=CtpU8mUAAAAJ&hl=en" target="_blank">Tian Fang<sup>1</sup></a>,
<a href="https://scholar.google.com.hk/citations?user=IXJm9jAAAAAJ&hl=en" target="_blank">David McKinnon<sup>1</sup></a>,
<a href="https://scholar.google.com.hk/citations?user=pa09Db8AAAAJ&hl=en" target="_blank">Yanghai Tsin<sup>1</sup></a>,
<a href="https://www.cse.ust.hk/~quan/" target="_blank">Long Quan<sup>3</sup></a> </h5>
<h5 style="color:#000000;"> <sup>1</sup>Apple, <sup>2</sup>Nanjing University, <sup>3</sup>HKUST </h5>
<br>
<p><a class="btn btn-secondary btn-lg" href="https://arxiv.org/abs/2303.17147" role="button">Paper</a> <a
class="btn btn-secondary btn-lg" href="https://github.com/apple/ml-neilfpp" role="button">Code</a> <a
class="btn btn-secondary btn-lg" href="https://apple.ent.box.com/s/2ulz3kjtca53aj3qlwjwyv27zfuug2x5" role="button">Data</a> </p>
</div>
</div>
</div>
</div>
</section>
<section>
<div class="container">
<div class="row">
<div class="col-12 text-center">
<h2>Abstract</h2>
<hr style="margin-top:0px">
<div class="b-row">
<div class="b-dics" style="width: 100%">
<img src="img/dtu_cam_normal_VolSDF.jpg" alt="VolSDF Normal" />
<img src="img/dtu_cam_normal_NeILF++.jpg" alt="Our Normal" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/dtu_cam_base_NeILF.jpg" alt="NeILF Base Color" />
<img src="img/dtu_cam_base_NeILF++.jpg" alt="Our Base Color" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/dtu_cam_render_NeILF.jpg" alt="NeILF Rendering" />
<img src="img/dtu_cam_render_NeILF++.jpg" alt="Our Rendering" />
</div>
</div>
<br>
<div class="b-row">
<div class="b-dics" style="width: 100%">
<img src="img/dtu_11_normal_VolSDF.jpg" alt="VolSDF Normal" />
<img src="img/dtu_11_normal_NeILF++.jpg" alt="Our Normal" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/dtu_11_base_NeILF.jpg" alt="NeILF Base Color" />
<img src="img/dtu_11_base_NeILF++.jpg" alt="Our Base Color" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/dtu_11_render_NeILF.jpg" alt="NeILF Rendering" />
<img src="img/dtu_11_render_NeILF++.jpg" alt="Our Rendering" />
</div>
</div>
<br>
<p class="text-left">We present a novel differentiable rendering framework for joint geometry,
material, and lighting estimation from multi-view images. In contrast to previous methods which
assume a simplified environment map or co-located flashlights, in this work, we formulate the
lighting of a static scene as one neural incident light field (NeILF) and one outgoing neural
radiance field (NeRF). The key insight of the proposed method is the union of the incident and
outgoing light fields through physically-based rendering and inter-reflections between surfaces,
making it possible to disentangle the scene geometry, material, and lighting from image
observations in a physically-based manner. The proposed incident light and inter-reflection
framework can be easily applied to other NeRF systems. We show that our method can not only
decompose the outgoing radiance into incident lights and surface materials, but also serve as a
surface refinement module that further improves the reconstruction detail of the neural surface.
We demonstrate on several datasets that the proposed method is able to achieve state-of-the-art
results in terms of the geometry reconstruction quality, material estimation accuracy, and the
fidelity of novel view rendering.
</p>
</div>
</div>
</div>
</section>
<br>
<br>
<section>
<div class="container">
<div class="row" style="margin-top:5px">
<div class="col-12 text-center">
<h2>Method Overview</h2>
<hr style="margin-top:0px">
</div>
</div>
<img class="img-fluid" src="img/pipeline.png" alt="">
<br>
<br>
<ul>
<li>A general light field representation by unifying the neural incident light field (<b>NeILF</b>) and
neural radiance field (<b>NeRF</b>) through <b>physically-based rendering</b> and
<b>inter-reflection</b> between surfaces.</li>
<li>An optimization scheme for the <b>joint estimation</b> of geometry, material, and lighting.</li>
<li>A real-world <b>linear HDR dataset</b> for material estimation and other neural rendering tasks.
</li>
<li>Can be easily applied to the NeRF family for <b>material decomposition</b> and <b>surface
refinement</b></li>
</ul>
</div>
</section>
<br>
<br>
<section>
<div class="container">
<div class="row">
<div class="col-12 text-center">
<h2>Inter-reflection</h2>
</div>
</div>
<hr style="margin-top:0px">
<div class="row">
<div class="col-12 text-center">
<div class="b-row">
<div class="b-dics" style="width: 100%">
<img src="img/syn_base_wo_ir.png" alt="Base. w/o IR" />
<img src="img/syn_base_w_ir.png" alt="Base. w. IR" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/syn_roughness_wo_ir.png" alt="Roug. w/o IR" />
<img src="img/syn_roughness_w_ir.png" alt="Roug. w. IR" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/syn_metallic_wo_ir.png" alt="Meta. w/o IR" />
<img src="img/syn_metallic_w_ir.png" alt="Meta. w. IR" />
</div>
</div>
<br>
<div class="b-row">
<div class="b-dics" style="width: 100%">
<img src="img/syn_neilf_wo_ir.png" alt="Incident Lights at point x1 w/o IR" />
<img src="img/syn_neilf_w_ir.png" alt="Incident Lights at point x1 w. IR" />
</div>
</div>
</div>
</div>
</div>
</section>
<br>
<br>
<section>
<div class="container">
<div class="row">
<div class="col-12 text-center">
<h2>HDR vs LDR</h2>
</div>
</div>
<hr style="margin-top:0px">
<div class="row">
<div class="col-12 text-center">
<p class="text-left">
As the commonly used LDR images are usually processed by unknown non-linear
tone-mapping, gamma correction, and value clipping, it may result in inaccurate
material and lighting estimation if we directly supervise the rendering value
with the LDR ground truth.
</p>
<p class="text-left">
Therefore, we constructed a real-world linear HDR dataset (by using Apple Internal tools)
for material estimation and neural rendering related tasks. Comparisons between
reconstructions using LDR and HDR images are visualized below.
</p>
<div class="b-row">
<div class="b-dics" style="width: 100%">
<img src="img/cat_base_ldr.png" alt="Base Color w. LDR" />
<img src="img/cat_base_hdr.png" alt="Base Color w. HDR" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/cat_roughness_ldr.png" alt="Roug. w. LDR" />
<img src="img/cat_roughness_hdr.png" alt="Roug. w. HDR" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/cat_metallic_ldr.png" alt="Meta. w. LDR" />
<img src="img/cat_metallic_hdr.png" alt="Meta. from HDR" />
</div>
</div>
<br>
<div class="b-row">
<div class="b-dics" style="width: 100%">
<img src="img/shoe_base_ldr.png" alt="Base. w. LDR" />
<img src="img/shoe_base_hdr.png" alt="Base. w. HDR" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/shoe_roughness_ldr.png" alt="Roug. w. LDR" />
<img src="img/shoe_roughness_hdr.png" alt="Roug. w. HDR" />
</div>
<div class="b-dics" style="width: 100%">
<img src="img/shoe_metallic_ldr.png" alt="Meta. w. LDR" />
<img src="img/shoe_metallic_hdr.png" alt="Meta. w. HDR" />
</div>
</div>
<br>
</div>
</div>
</div>
</section>
<br>
<br>
<section>
<div class="container">
<div class="row">
<div class="col-12 text-center">
<h2>Relighting</h2>
</div>
</div>
<hr style="margin-top:0px">
<!-- <div class="b-row">
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/dtu_scan110_geometry0001-0060.mp4" type="video/mp4">
</video>
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/dtu_scan110_city0001-0060.mp4" type="video/mp4">
</video>
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/dtu_scan110_studio0001-0060.mp4" type="video/mp4">
</video>
</div>
<br> -->
<div class="b-row">
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/hdr_lucky_cat_geometry0001-0125.mp4" type="video/mp4">
</video>
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/hdr_lucky_cat_city0001-0125.mp4" type="video/mp4">
</video>
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/hdr_lucky_cat_studio0001-0125.mp4" type="video/mp4">
</video>
</div>
<div class="b-row">
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/hdr_typewritter_geometry0001-0125.mp4" type="video/mp4">
</video>
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/hdr_typewritter_city0001-0125.mp4" type="video/mp4">
</video>
<video width="33%" playsinline="" autoplay="" loop="" preload="" muted="">
<source src="img/videos/hdr_typewritter_studio0001-0125.mp4" type="video/mp4">
</video>
</div>
</div>
</section>
<br>
<br>
<footer class="text-center">
<div class="container">
<div class="row ">
<div class="col-12">
<h2>Citation</h2>
<hr style="margin-top:0px">
<div class="bibtexsection">
<pre><code>
@article{zhang2023neilf++,
title={NeILF++: Inter-reflectable Light Fields for Geometry and Material Estimation},
author={Zhang, Jingyang and Yao, Yao and Li, Shiwei and Liu, Jingbo and Fang, Tian and McKinnon,
David and Tsin, Yanghai and Quan, Long},
journal={International Conference on Computer Vision (ICCV)},
year={2023}
}
</code>
</div>
</div>
</div>
</div>
</footer>
<br>
<div class="row">
<div class="col-12 text-center">
<p>
Acknowledgements: The website template was borrowed from <a href="https://lioryariv.github.io/idr/">Lior Yariv</a>.
Image sliders are based on <a href="https://github.com/abelcabezaroman/definitive-image-comparison-slider">dics</a>.
</p>
</div>
</div>
<br>
<br>
<!-- jQuery (necessary for Bootstrap's JavaScript plugins) -->
<script src="js/jquery-3.4.1.min.js"></script>
<!-- Include all compiled plugins (below), or include individual files as needed -->
<script src="js/popper.min.js"></script>
<script src="js/bootstrap-4.4.1.js"></script>
</body>
</html>