1<!DOCTYPE html> 2<html lang="en"> 3 <head> 4 <meta charset="utf-8"/> 5 <meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no"/> 6 <meta name="description" content=""/> 7 <meta name="author" content=""/> 8 <title>Zeping Ren</title> 9 <link rel="icon" type="image/x-icon" href="assets/apple.ico"/> 10 <!-- Font Awesome icons (free version)--> 11 <!-- <link rel="stylesheet" href="./css/font-awesome.min.css"/> --> 12
12<script src="https://use.fontawesome.com/releases/v5.15.4/js/all.js" crossorigin="anonymous"></script>
12 13 <!-- Google fonts--> 14 <link href="https://fonts.googleapis.com/css?family=Saira+Extra+Condensed:500,700" rel="stylesheet" type="text/css"/> 15 <link href="https://fonts.googleapis.com/css?family=Muli:400,400i,800,800i" rel="stylesheet" type="text/css"/> 16 <!-- Core theme CSS (includes Bootstrap)--> 17 <link href="css/styles.css" rel="stylesheet"/> 18 </head> 19 <body id="page-top"> 20 <!-- Navigation--> 21 <nav class="navbar navbar-expand-lg navbar-dark bg-primary fixed-top" id="sideNav"> 22 <a class="navbar-brand js-scroll-trigger" href="#page-top"> 23 <span class="d-none d-lg-block"> 24 <img class="img-fluid img-profile rounded-circle mx-auto mb-2" src="assets/me.jpg" alt="..."/> 25 <h3 class="mb-0">Zeping Ren</h3> 26 </span> 27 </a> 28 <button class="navbar-toggler" type="button" data-bs-toggle="collapse" data-bs-target="#navbarResponsive" aria-controls="navbarResponsive" aria-expanded="false" aria-label="Toggle navigation"> 29 <span class="navbar-toggler-icon"></span> 30 </button> 31 <div class="collapse navbar-collapse" id="navbarResponsive"> 32 <ul class="navbar-nav"> 33 <li class="nav-item"> 34 <a class="nav-link js-scroll-trigger" href="#about"> 35 <i class="fa fa-home" aria-hidden="true"></i> 36 About 37 </a> 38 </li> 39 <!-- fa fa-ellipsis-h--> 40 <li class="nav-item"> 41 <a class="nav-link js-scroll-trigger" href="#research"> 42 <i class="fa fa-flask" aria-hidden="true"></i> 43 Research 44 </a> 45 </li> 46 <li class="nav-item"> 47 <a class="nav-link js-scroll-trigger" href="#more"> 48 <i class="fa fa-graduation-cap" aria-hidden="true"></i> 49 More 50 </a> 51 </li> 52 </ul> 53 </div> 54 </nav> 55 <!-- Page Content--> 56 <div class="container-fluid p-0"> 57 <!-- About--> 58 <section class="resume-section" id="about"> 59 <div class="resume-section-content"> 60 <!--Only show the image on mobile--> 61 <span class="d-lg-none"> 62 <img class="img-fluid img-profile rounded-circle mx-auto mb-2" width=150px src="assets/me.jpg" alt="..."/> 63 </span> 64 <h2>Zeping Ren</h2> 65 <div class="subheading mb-3"> 66 <span style="text-decoration:none; color:inherit;"> 67 Master Student, Tsinghua University 68 </span> 69 <br> 70 <span style="text-decoration:none; color:inherit;"> 71 Hometown: Chongqing,China 72 </span> 73 <br> 74 <span style="font-size:1.1rem"> 75 Email:[email protected] 76 </span> 77 </div> 78 <!-- <div class="social-icons"> 79 <a class="social-icon" href="mailto:[email protected]"> 80 <i class="fa fa-envelope"></i> 81 </a> 82 <a class="social-icon" href="https://scholar.google.fr/citations?user=ceSzF9YAAAAJ&hl=en" class="text-info" title="Google Scholar"> 83 <i class="ai ai-google-scholar"></i> 84 </a> 85 <a class="social-icon" href="https://github.com/gulvarol"> 86 <i class="fa fa-github"></i> 87 </a> 88 <a class="social-icon" href="https://www.linkedin.com/in/gulvarol/"> 89 <i class="fa fa-linkedin"></i> 90 </a> 91 <a class="social-icon" href="https://twitter.com/gulvarol"> 92 <i class="fa fa-twitter"></i> 93 </a> 94 </div> --> 95 <p> 96 <br> 97 <b>Bio:</b> 98 I am a master student studied in Tsinghua Shenzhen International Graduate School, 99 advised by <a href="https://www.sigs.tsinghua.edu.cn/lx/main.htm">Prof.Xiu Li</a>. 100 I got my undergraduate degree in Department of Automation, Tsinghua University, 101 under the supervision of <a href="http://liuyebin.com/">Prof.Yebin Liu</a>. 102 My research is focus on 3D pose estimation and motion generation. 103 </p> 104 </div> 105 </section> 106 107 </div> 108 <hr class="m-0"/> 109 <!--Research--> 110 <section class="resume-section" id="research"> 111 <div class="resume-section-content"> 112 <h3 class="mb-3">Research</h3> 113 <!-- crossdiff --> 114 <div class="row"> 115 <div class="col-xs-10 col-sm-4 col-md-4"> 116 <a href="https://wonderno.github.io/CrossDiff-webpage/"> 117 <img class="img-thumbnail mb-3" src="./assets/crossdiff.jpg" alt="Realistic Human Motion Generation with Cross-Diffusion Models"> 118 </a> 119 </div> 120 <div class="col-xs-12 col-sm-8 col-md-8"> 121 <strong>Realistic Human Motion Generation with Cross-Diffusion Models</strong> 122 <br> 123 <u>Zeping Ren</u>, Shaoli Huang, Xiu Li 124 <br> 125 <em>ECCV </em> 126 2024.<br> 127 <a href="https://arxiv.org/abs/2312.10993"> 128 <button type="button" class="btn btn-primary btn-sm"> pdf </button></a> 129 <a href="https://wonderno.github.io/CrossDiff-webpage/"> 130 <button type="button" class="btn btn-primary btn-sm">project page</button></a> 131 <button type="button" class="btn btn-primary btn-sm" data-bs-toggle="collapse" data-bs-target="#bibtex1">bibtex</button> 132 <div id="bibtex1" class="collapse"> 133 <pre> 134<tt>@article{ren2023realistic, 135 title={Realistic Human Motion Generation with Cross-Diffusion Models}, 136 author={Ren, Zeping and Huang, Shaoli and Li, Xiu}, 137 journal={arXiv preprint arXiv:2312.10993}, 138 year={2023} 139}</tt> 140 </pre> 141 </div> 142 <button type="button" class="btn btn-primary btn-sm" data-bs-toggle="collapse" data-bs-target="#abstract3">abstract</button> 143 <div id="abstract3" class="collapse"> 144 <p class="bg-light"> 145 We introduce the Cross Human Motion Diffusion Model (CrossDiff), a novel approach for generating high-quality human motion based on textual descriptions. 146 Our method integrates 3D and 2D information using a shared transformer network within the training of the diffusion model, unifying motion noise into a single feature space. 147 This enables cross-decoding of features into both 3D and 2D motion representations, regardless of their original dimension. 148 The primary advantage of CrossDiff is its cross-diffusion mechanism, which allows the model to reverse either 2D or 3D noise into clean motion during training. 149 This capability leverages the complementary information in both motion representations, capturing intricate human movement details often missed by models relying solely on 3D information. 150 Consequently, CrossDiff effectively comb
150ines the strengths of both representations to generate more realistic motion sequences. 151 In our experiments, our model demonstrates competitive state-of-the-art performance on text-to-motion benchmarks. 152 Moreover, our method consistently provides enhanced motion generation quality, capturing complex full-body movement intricacies. 153 Additionally, our approach accommodates using 2D motion data without 3D motion ground truth during training to generate 3D motion, highlighting its potential for broader applications and efficient use of available data resources. 154 155 </p> 156 </div> 157 <a href="https://github.com/wonderNo/crossdiff"> 158 <button type="button" class="btn btn-primary btn-sm">code </button> 159 </a> 160 <div style="height:30px;"></div> 161 </div> 162 </div> 163 <!-- end of crossdiff --> 164 165 <!-- finedance --> 166 <div class="row"> 167 <div class="col-xs-10 col-sm-4 col-md-4"> 168 <a href="https://li-ronghui.github.io/finedance"> 169 <img class="img-thumbnail mb-3" src="./assets/finedance.png" alt="FineDance: A Fine-grained Choreography Dataset for 3D Full Body Dance Generation"> 170 </a> 171 </div> 172 <div class="col-xs-12 col-sm-8 col-md-8"> 173 <strong>FineDance: A Fine-grained Choreography Dataset for 3D Full Body Dance Generation</strong> 174 <br> 175 Ronghui Li*, Junfan Zhao*, Yachao Zhang, Mingyang Su, <u>Zeping Ren</u>, Han Zhang, Yansong Tang, Xiu Li 176 <br> 177 <em>ICCV </em> 178 2023.<br> 179 <a href="https://arxiv.org/abs/2212.03741v4"> 180 <button type="button" class="btn btn-primary btn-sm"> pdf </button></a> 181 <a href="https://li-ronghui.github.io/finedance"> 182 <button type="button" class="btn btn-primary btn-sm"> project page </button></a> 183 <button type="button" class="btn btn-primary btn-sm" data-bs-toggle="collapse" data-bs-target="#bibtex2">bibtex</button> 184 <div id="bibtex2" class="collapse"> 185 <pre> 186<tt>@InProceedings{Li_2023_ICCV, 187 author = {Li, Ronghui and Zhao, Junfan and Zhang, Yachao and Su, Mingyang and Ren, Zeping and Zhang, Han and Tang, Yansong and Li, Xiu}, 188 title = {FineDance: A Fine-grained Choreography Dataset for 3D Full Body Dance Generation}, 189 booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)}, 190 month = {October}, 191 year = {2023}, 192 pages = {10234-10243} 193}</tt> 194 </pre> 195 </div> 196 <button type="button" class="btn btn-primary btn-sm" data-bs-toggle="collapse" data-bs-target="#abstract2">abstract</button> 197 <div id="abstract2" class="collapse"> 198 <p class="bg-light"> 199 Generating full-body and multi-genre dance sequences from given music is a challenging task, due to the limitations of existing datasets and the inherent complexity of the fine-grained hand motion and dance genres. To address these problems, we propose FineDance, which contains 14.6 hours of music-dance paired data, with fine-grained hand motions, fine-grained genres (22 dance genres), and accurate posture. To the best of our knowledge, FineDance is the largest music-dance paired dataset with the most dance genres. Additionally, to address monotonous and unnatural hand movements existing in previous methods, we propose a full-body dance generation network, which utilizes the diverse generation capabilities of the diffusion model to solve monotonous problems, and use expert nets to solve unreal problems. To further enhance the genrematching and long-term stability of generated dances, we propose a Genre&Coherent aware Retrieval Module. Besides, we propose a new metric named Genre Matching Score to measure the genre matching between dance and music. Quantitative and qualitative experiments demonstrate the quality of FineDance, and the state-of-the-art performance of FineNet. 200 201 </p> 202 </div> 203 <a href="https://github.com/li-ronghui/FineDance"> 204 <button type="button" class="btn btn-primary btn-sm">
204code </button> 205 </a> 206 <div style="height:30px;"></div> 207 </div> 208 </div> 209 <!-- end of finedance --> 210 211 <!-- realtime_capture --> 212 <div class="row"> 213 <div class="col-xs-10 col-sm-4 col-md-4"> 214 <a href="http://www.liuyebin.com/realtime_cap_project/realtotalcap/realtotalcap.html"> 215 <img class="img-thumbnail mb-3" src="./assets/realtotalcap.jpg" alt="Real-time Sparse-view Multi-person Total Motion Capture"> 216 </a> 217 </div> 218 <div class="col-xs-12 col-sm-8 col-md-8"> 219 <strong>Real-time Sparse-view Multi-person Total Motion Capture</strong> 220 <br> 221 Yuxiang Zhang, <u>Zeping Ren</u>, Liang An, Hongwen Zhang, Tao Yu, Yebin Liu 222 <br> 223 <!-- <em>arXiv </em> --> 224 2022.<br> 225 <a href="http://www.liuyebin.com/realtime_cap_project/realtotalcap/realtotalcap.html"> 226 <button type="button" class="btn btn-primary btn-sm"> project page </button></a> 227 <button type="button" class="btn btn-primary btn-sm" data-bs-toggle="collapse" data-bs-target="#abstract1">abstract</button> 228 <div id="abstract1" class="collapse"> 229 <p class="bg-light"> 230 Real-time multi-person total motion capture is one of the most challenging tasks for human motion capture. The multi-view configuration reduces occlusions and depth ambiguity yet further complicates the problem by introducing cross-view association into consideration. In this paper, we contribute the first real-time multi-person total motion capture system under sparse views. To enable full body cross-view association in real-time, we propose a highly efficient association algorithm, named Clique Unfolding, by reformulating the widely used fast unfolding algorithm for community detection. Moreover, an adaptive motion prior based on human motion prediction is proposed to improve the SMPL-X fitting performance in the final step. Benefiting from the proposed association and fitting methods, our system achieves robust, efficient and accurate multi-person total motion capture results. Experiments and results demonstrate the efficiency and effectiveness of the proposed method. 231 </p> 232 </div> 233 <div style="height:30px;"></div> 234 </div> 235 </div> 236 <!-- end of realtime_capture --> 237 238 </div> 239 </section> 240 <!-- end of Research --> 241 242 <hr class="m-0"/> 243 <!-- More--> 244 <section class="resume-section" id="more"> 245 <div class="resume-section-content"> 246 <h3 class="mb-3">More</h3> 247 <p> 248 I'm not a magician, but I never stop creating miracles. 249 </p> 250 </div> 251 </section> 252 <!-- end of More --> 253 254 255 <!-- Bootstrap core JS--> 256
256<script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/js/bootstrap.bundle.min.js"></script>
256 257 <!-- Core theme JS--> 258
258<script src="js/scripts.js"></script>
258 259 </body> 260</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.