1<!DOCTYPE html> 2 3<head> 4 <meta charset="utf-8" /> 5 <title>DiffSound</title> 6 <meta content="DiffSound." name="description" /> 7 <meta content="DiffSound Networks" property="og:title" /> 8 <meta content="DiffSound" property="og:description" /> 9 <meta content="width=device-width, initial-scale=1" name="viewport" /> 10 <link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/css/bootstrap.min.css" 11 integrity="sha384-Gn5384xqQ1aoWXA+058RXPxPg6fy4IWvTNh0E263XmFcJlSAwiGgFAW/dAiS6JXm" crossorigin="anonymous"> 12
12<script src="./google.js" type="text/javascript"></script>
12 13
13<script type="text/javascript">WebFont.load({ 14 google: { 15 families: ["Lato:100,100italic,300,300italic,400,400italic,700,700italic,900,900italic", "Montserrat:100,100italic,200,200italic,300,300italic,400,400italic,500,500italic,600,600italic,700,700italic,800,800italic,900,900italic", "Ubuntu:300,300italic,400,400italic,500,500italic,700,700italic", "Changa One:400,400italic", "Open Sans:300,300italic,400,400italic,600,600italic,700,700italic,800,800italic", "Varela Round:400", "Bungee Shade:regular", "Roboto:300,regular,500" 16 , "Radio Canada", "Source Sans Pro", "Roboto Slab"] 17 } 18 });</script>
18 19 20 <style> 21 body { 22 font-family: 'Microsoft Yahei' !important; 23 text-align: center; 24 } 25 26 .all-wrap { 27 margin: 0 auto; 28 width: 65%; 29 } 30 31 .title { 32 font-family: 'Radio Canada'; 33 color: #3c3c3c; 34 font-weight: 800; 35 font-size: 3vw; 36 line-height: 1.3; 37 margin-top: 2vw; 38 } 39 40 .subheader { 41 font-family: 'Open Sans', sans-serif; 42 color: #333; 43 font-weight: 700; 44 font-size: 1.5vw; 45 margin-top: 1.3vw; 46 margin-bottom: 1.5vw; 47 } 48 49 .affiliation { 50 display: inline; 51 font-family: Ubuntu, Helvetica, sans-serif; 52 color: #333; 53 font-size: 1.2vw; 54 font-weight: 400; 55 letter-spacing: 0; 56 line-height: 1.8vw; 57 } 58 59 .teaser { 60 max-width: 100%; 61 } 62 63 .container-row { 64 width: 100%; 65 box-sizing: border-box; 66 } 67 68 .icon-img-container { 69 position: relative; 70 float: left; 71 width: 33.333%; 72 box-sizing: border-box; 73 margin-top: 1.6vw; 74 } 75 76 .icon-text-container { 77 position: relative; 78 float: left; 79 width: 33.333%; 80 box-sizing: border-box; 81 margin-bottom: 1.6vw; 82 } 83 84 .icon-img { 85 width: 25%; 86 filter: contrast(88%); 87 opacity: .77; 88 } 89 90 .icon-text { 91 font-family: 'Open Sans', sans-serif; 92 font-size: 1.2vw; 93 font-weight: 550; 94 line-height: 2vw; 95 } 96 97 .section-title { 98 text-transform: uppercase; 99 font-family: 'Radio Canada'; 100 color: #3c3c3c; 101 font-size: 2vw; 102 font-weight: 800; 103 line-height: 1.3; 104 105 } 106 107 .long-text { 108 font-family: 'Source Sans Pro'; 109 font-size: 1.3vw; 110 font-weight: 500; 111 line-height: 1.2; 112 text-align: justify; 113 } 114 115 .grey_container { 116 background-color: #f2f2f2; 117 } 118 119 .bibtex { 120 text-align: left; 121 margin-left: 2vw; 122 font-size: larger; 123 font-size: 1.2vw; 124 } 125 </style> 126</head> 127 128<body> 129 <div class="all-wrap"> 130 <h1 class="title">DiffSound: Differentiable Modal Sound Rendering and Inverse Rendering for Diverse Inference Tasks 131 </h1> 132 <h1 class="subheader">SIGGRAPH 2024</h1> 133 <div> 134 <h1 class="affiliation" style="margin-right: 20px;">Xutong Jin*<sup>1</sup></h1> 135 <h1 class="affiliation" style="margin-right: 20px;">Chenxi Xu*<sup>1</sup></h1> 136 <h1 class="affiliation" style="margin-right: 20px;">Ruohan Gao<sup>2</sup></h1> 137 <h1 class="affiliation" style="margin-right: 20px;">Jiajun Wu<sup>3</sup></h1> 138 <h1 class="affiliation" style="margin-right: 20px;">Guoping Wang<sup>1</sup></h1> 139 <h1 class="affiliation" style="margin-right: 20px;"><a 140 href="http://www.graphics.pku.edu.cn/members/lisheng/index.htm"> Sheng Li**</a> <sup>1</sup></h1> 141 </div> 142 143 <div style="margin-top: 0.4vw;margin-bottom: 1vw;"> 144 <h1 class="affiliation" style="margin-right: 40px"><sup>1</sup>Peking University</h1> 145 <h1 class="affiliation" style="margin-right: 40px"><sup>2</sup>University of Maryland, College Park </h1> 146 <h1 class="affiliation"><sup>3</sup>Stanford University</h1> 147 </div> 148 149 <div style="margin-top: 0.4vw;margin-bottom: 1vw;"> 150 151 <h1 class="affiliation" style="margin-right: 40px"><sup>*</sup>Equal contributions</h1> 152 <h1 class="affiliation"><sup>**</sup>Corresponding author</h1> 153 154 </div> 155 156 157 158 <div> 159 <span class="center"><img class="teaser" src="images/teaser.png"></span> 160 </div> 161 162 <div class="container-row"> 163 <div class="icon-img-container"> 164 <a href="./Paper.pdf"> 165 <img src="./images/paper.png" alt="paper" class="icon-img" /> 166 </a> 167 </div> 168 <div class="icon-img-container"> 169 <a href="https://github.com/TechnetiumMan/DiffSound"> 170 <img src="./images/github.png" alt="github" class="icon-img" /> 171 </a> 172 </div> 173 <div class="icon-img-container"> 174 <a 175 href="./DiffSound.mp4"> 176 <img src="./images/video.png" alt="github" class="icon-img" /> 177 </a> 178 </div> 179 </div> 180 181 <div class="container-row"> 182 <div class="icon-text-container"> 183 <strong class="icon-text">Paper</strong> 184 </div> 185 <div class="icon-text-container"> 186 <strong class="icon-text">Github Code</strong> 187 </div> 188 <div class="icon-text-container"> 189 <strong class="icon-text">Supplementary Video</strong> 190 </div> 191 </div> 192 193 194 <div class="container-row"> 195 <video loop="" playsinline="" controls="" width="100%"> 196 <source 197 src="./DiffSound.mp4" 198 type="video/mp4"> 199 </video> 200 </div> 201 <div class="container-row"> 202 <video loop="" playsinline="" controls="" width="100%"> 203 <source 204 src="./fast-forward.mp4" 205 type="video/mp4"> 206 </video> 207 </div> 208 209 <div class="w-container"> 210 <h2 class="section-title">Abstract</h2> 211 <p class="long-text">Accurately estimating and simulating the physical properties of objects from real-world sound 212 recordings is of great practical importance in the fields of vision, graphics, and robotics. However, the 213 progress in these directions has been limited---prior differentiable rigid or soft body simulation techniques 214 cannot be directly applied to modal sound synthesis due to the high sampling rate of audio, while previous audio 215 synthesizers often do not fully model the accurate physical properties of the sounding objects. We propose 216 DiffSound, a differentiable sound rendering framework for physics-based modal sound synthesis, which is based on 217 an implicit shape representation, a new high-order finite element analysis module, and a differentiable audio 218 synthesizer. Our framework can solve a wide range of inverse problems thanks to the differentiability of the 219 entire pipeline, including physical parameter estimation, geometric shape reasoning, and impact position 220 prediction. Experimental results demonstrate the effectiveness of our approach, highlighting its ability to 221 accurately reproduce the target sound in a physics-based manner. DiffSound serves as a valuable tool for various 222 sound synthesis and analysis applications. 223 </p> 224 </div> 225 <hr> 226 227 <h2 class="section-title">Bibtex</h2> 228 <div class="grey_container"> 229 <div class="bibtex"> 230 <pre><code> 231@inproceedings{10.1145/3641519.3657493, 232 author = {Jin, Xutong and Xu, Chenxi and Gao, Ruohan and Wu, Jiajun and Wang, Guoping and Li, Sheng}, 233 title = {DiffSound: Differentiable Modal Sound Rendering and Inverse Rendering for Diverse Inference Tasks}, 234 year = {2024}, 235 isbn = {9798400705250}, 236 publisher = {Association for Computing Machinery}, 237 address = {New York, NY, USA}, 238 url = {https://doi.org/10.1145/3641519.3657493}, 239 doi = {10.1145/3641519.3657493}, 240 booktitle = {ACM SIGGRAPH 2024 Conference Papers}, 241 articleno = {128}, 242 numpages = {12}, 243 keywords = {audio, differentiable simulation, modal analysis, sound synthesis, vibration}, 244 location = {Denver, CO, USA}, 245 series = {SIGGRAPH '24} 246} 247 </code></pre> 248 </div> 249 </div> 250 251 </div> 252 253</body> 254 255</html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.