1<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"> 2<html xmlns="http://www.w3.org/1999/xhtml"> 3 <head> 4 <meta charset="UTF-8"> 5 <meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no"> 6 <meta http-equiv="x-ua-compatible" content="ie=edge"> 7 8 <meta http-equiv="content-type" content="text/html; charset=utf-8" /> 9 <meta name="description" content="" /> 10 <meta name="keywords" content="" /> 11 <title>Haoqi Fan</title> 12 <link rel="stylesheet" type="text/css" href="css/style.css" /> 13 14 15 <link rel="stylesheet" href="css/bootstrap.min.css"> 16 <link rel="stylesheet" href="css/academicons.min.css"> 17 <link rel="stylesheet" href="https://cdn.rawgit.com/jpswalsh/academicons/master/css/academicons.min.css"> 18 <link rel="stylesheet" href="css/font-awesome.min.css"> 19 20 21 <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/5.11.2/css/all.min.css" integrity="sha256-+N4/V/SbAFiW1MPBCXnfnP9QSN3+Keu+NlB+0ev/YKQ=" crossorigin="anonymous"> 22 <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/fancybox/3.5.7/jquery.fancybox.min.css" integrity="sha256-Vzbj7sDDS/woiFS3uNKo8eIuni59rjyNGtXfstRzStA=" crossorigin="anonymous"> 23 <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.15.10/styles/dracula.min.css" crossorigin="anonymous" title="hl-dark" disabled> 24
24<script src="https://cdnjs.cloudflare.com/ajax/libs/lazysizes/5.1.2/lazysizes.min.js" integrity="sha256-Md1qLToewPeKjfAHU1zyPwOutccPAm5tahnaw7Osw0A=" crossorigin="anonymous" async></script>
24 25 <link rel="stylesheet" href="https://fonts.googleapis.com/css?family=Libre+Baskerville%7CMontserrat:400,700%7CRoboto%7CRoboto Mono&display=swap"> 26 <link rel="stylesheet" href="/css/academic.css"> 27 <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.15.10/styles/github.min.css" crossorigin="anonymous" title="hl-light"> 28 <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/academicons/1.8.6/css/academicons.min.css" integrity="sha256-uFVgMKfistnJAfoCUQigIl+JfUaP47GrRKjf6CTPVmw=" crossorigin="anonymous"> 29 30 <style> 31 a.plain-black { 32 text-decoration: none; 33 color: black; 34 margin: 5px; 35 } 36 </style> 37 </head> 38 <body> 39 40 41 42 <div id="wrapper"> 43 44 <div id="inner"> 45 <div id="page"> 46 <div id="content"> 47 <table> 48 <tr> 49 <td align="left" width = "180"><img class="image" src="image/haoqi_fan.jpg" width = "135px" alt="" /></td> <td> 50 <h2>Haoqi Fan</h2> 51 <p> Researcher in Computer Vision <br> 52 <a href="https://ai.facebook.com/" target="_blank">Facebook AI Research</a> <br> 53 E-mail: haoqifan [at] gmail [dot] com <br> 54 <a class="plain-black" href="mailto:[email protected]"><i class="fa fa-2x fa-envelope"></i></a> 55 <a class="plain-black" href="https://scholar.google.com/citations?user=76B8lrgAAAAJ&hl=en" target="_blank"><i class="ai ai-google-scholar-square ai-2x"></i></a> 56 </p> 57 </td> 58 <td> 59 </tr> 60 </table> 61 <br/> 62 <h3>About Me</h3> 63 <table> 64 <tr> 65 <td align="left" bgcolor="" width = "1000px", line-height = "120px" > 66 <p>Haoqi Fan is an AI Researcher who works on multimodal foundational models. He spent 7 years at <a href="https://ai.facebook.com/" target="_blank">Facebook AI Research (FAIR)</a>. He graduated from <a href = "https://www.ri.cmu.edu/" target="_blank">the Robotics Institute</a> of Carnegie Mellon University. His research interests lie in computer vision and deep learning. 67 </p> 68 </td> 69 </tr> 70 </table> 71 72 <h3>Recent News</h3> 73 74 75 <! ââ BAGEL ââ> 76 <table> 77 <tr> 78 <td align="left" bgcolor="" width = "1000px" > 79 <li> <b>May 2025</b>: Released <a href="https://bagel-ai.org/">BAGEL</a> for unified multimodal research and application, it quickly soared to the #1 trending model on Hugging Face. <a href="https://github.com/bytedance-seed/BAGEL"><img alt="GitHub stars" src="https://img.shields.io/github/stars/bytedance-seed/BAGEL?style=social"></a></li> 80 <br/> 81 </td> 82 </tr> 83 </table> 84 85 86 87 <! ââ PTV ââ> 88 <table> 89 <tr> 90 <td align="left" bgcolor="" width = "1000px" > 91 <li> <b>Apr 2021</b>: Released <a href = "https://pytorchvideo.org/" target="_blank">PyTorchVideo</a> - a deep learning library for video understanding research. 92 <a href="https://github.com/facebookresearch/pytorchvideo"><img alt="GitHub stars" src="https://img.shields.io/github/stars/facebookresearch/pytorchvideo?style=social"></a> 93 </li> 94 <br/> 95 </td> 96 </tr> 97 </table> 98 99 100 101 <! ââ ICCV Tutorialââ> 102 <table> 103 <tr> 104 <td align="left" bgcolor="" width = "1000px" > 105 <li> <b>Jun 2020</b>: Co-organized tutorials on Visual Recognition for Images, Video, and 3D at <a href = "http://s9xie.github.io/Tutorials/CVPR2020/" target="_blank">CVPR 2020</a> and <a href = "http://s9xie.github.io/Tutorials/ECCV2020/" target="_blank">ECCV 2020</a>.</li> 106 <br/> 107 </td> 108 </tr> 109 </table> 110 111 112 113 <! ââ PySlowFast Releaseââ> 114 <table> 115 <tr> 116 <td align="left" bgcolor="" width = "1000px" > 117 <li> <b>Nov 2019</b>: Released <a href="https://github.com/facebookresearch/SlowFast">PySlowFast</a> codebase for video understanding research. <a href="https://github.com/facebookresearch/SlowFast"><img alt="GitHub stars" src="https://img.shields.io/github/stars/facebookresearch/SlowFast?style=social"></a></li> 118 <br/> 119 </td> 120 </tr> 121 </table> 122 123 <! ââ ICCV Tutorialââ> 124 <table> 125 <tr> 126 <td align="left" bgcolor="" width = "1000px" > 127 <li> <b>Nov 2019</b>: Co-organized a tutorial on <a href = "https://alexander-kirillov.github.io/tutorials/visual-recognition-iccv19/" target="_blank">Images, Video, and 3D</a> research and code at ICCV 2019.</li> 128 <br/> 129 </td> 130 </tr> 131 </table> 132 133 <! ââ AVA Winââ> 134 <table> 135 <tr> 136 <td align="left" bgcolor="" width = "1000px" > 137 <li> <b>June 2019</b>: Won the 1st place of <a href = "http://research.google.com/ava/challenge.html" target="_blank">AVA video activity detection challenge</a> at the <a href = "http://activity-net.org/challenges/2019/" target="_blank">International Challenge on Activity Recognition (ActvityNet) ICCV 2019</a>.</li> 138 <br/> 139 </td> 140 </tr> 141 </table> 142 143 <!-- 144 <! ââ Outstanding Reviewerââ> 145 <table> 146 <tr> 147 <td align="left" bgcolor="" width = "1000px" > 148 <li> <b>June 2019</b>: Selected as CVPR 2019 <a href = "https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8954223" target="_blank">Outstanding Reviewer</a>.</li> 149 <br/> 150 </td> 151 </tr> 152 </table> 153 --> 154 155 156 <h3>Selected Publications</h3> 157 158 <! ââ BAGEL ââ> 159 <table> 160 <tr> 161 <td align="left" width = "50px"></td> 162 <td align="left" width = "350px"> 163 <img class="image" src="image/bagel-promo-vid.gif" width = "250px" alt="" /> 164 </td> 165 <td align="left" bgcolor="" width = "600px" > 166 167 <b>Emerging Properties in Unified Multimodal Pretraining</b> <br/> 168 Chaorui Dengâ, Deyao Zhuâ, Kunchang Liâ, Chenhui Gouâ, Feng Liâ, 169 Zeyu Wang, Shu Zhong, Weihao Yu, Xiaonan Nie, Ziang Song, Guang Shi§, 170 <b>Haoqi Fanââ 171 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2505.14683" target="_blank" rel="noopener"> 172 <i class="fas fa-file-pdf mr-1"></i> Paper 173 </a> 174 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://bagel-ai.org/" target="_blank" rel="noopener"> 175 <i class="fas fa-code mr-1"></i> 176 website 177 </a> 178 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/bytedance-seed/BAGEL" target="_blank" rel="noopener"> 179 <i class="fas fa-code mr-1"></i> 180 code 181 </a> 182 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://huggingface.co/ByteDance-Seed/BAGEL-7B-MoT" target="_blank" rel="noopener"> 183 <i class="fas fa-code mr-1"></i> 184 huggingface 185 </a> 186 <br/> 187 </td> 188 </tr> 189 </table> 190 191 192 <! ââ FLIP ââ> 193 <table> 194 <tr> 195 <td align="left" width = "50px"></td> 196 <td align="left" width = "350px"> 197 <img class="image" src="image/flip.png" width = "250px" alt="" /> 198 </td> 199 <td align="left" bgcolor="" width = "600px" > 200 201 <b>Scaling Language-Image Pre-training via Masking</b> <br/> 202 Yanghao Li*, <b>Haoqi Fan</b>*, Ronghang Hu*, Christoph Feichtenhoferâ , Kaiming Heâ <br/> 203 <i>Arxiv</i>, 2022<br/> 204 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2212.00794" target="_blank" rel="noopener"> 205 <i class="fas fa-file-pdf mr-1"></i> Paper 206 </a> 207 <br/> 208 209 </td> 210 </tr> 211 </table> 212 213 214 <! ââ Video MAE ââ> 215 <table> 216 <tr> 217 <td align="left" width = "50px"></td> 218 <td align="left" width = "350px"> 219 <img class="image" src="image/mae_st.png" width = "250px" alt="" /> 220 </td> 221 <td align="left" bgcolor="" width = "600px" > 222 223 <b>Masked Autoencoders As Spatiotemporal Learners</b> <br/> 224 Christoph Feichtenhofer*, <b>Haoqi Fan</b>*, Yanghao Li, Kaiming He<br/> 225 <i>NeurIPS</i>, 2022<br/> 226 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2205.09113" target="_blank" rel="noopener"> 227 <i class="fas fa-file-pdf mr-1"></i> Paper 228 </a> 229 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/mae_st" target="_blank" rel="noopener"> 230 <i class="fas fa-code mr-1"></i> 231 code 232 </a> 233 <br/> 234 235 </td> 236 </tr> 237 </table> 238 239 240 241 <! ââ MViT ââ> 242 <table> 243 <tr> 244 <td align="left" width = "50px"></td> 245 <td align="left" width = "350px"> 246 <img class="image" src="image/mvit-img-det-vid.png" width = "250px" alt="" /> 247 </td> 248 <td align="left" bgcolor="" width = "600px" > 249 250 <b>Multiscale Vision Transformers</b> <br/> 251 <b>Haoqi Fan*</b>, Bo Xiong*, Karttikeya Mangalam*, Yanghao Li*, Zhicheng Yan, Jitendra Malik, Christoph Feichtenhofer<br/> 252 <i>ICCV</i>, 2021<br/> 253 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2104.11227" target="_blank" rel="noopener"> 254 <i class="fas fa-file-pdf mr-1"></i> Paper 255 </a> 256 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/SlowFast" target="_blank" rel="noopener"> 257 <i class="fas fa-code mr-1"></i> 258 code 259 </a> 260 <br/> 261 262 <b>Improved Multiscale Vision Transformers for Classification and Detection</b> <br/> 263 Yanghao Li, Chao-Yuan Wu, <b>Haoqi Fan</b>, Karttikeya Mangalam, Bo Xiong, Jitendra Malik, Christoph Feichtenhofer<br/> 264 <i>Tech Report</i>, 2021<br/> 265 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2112.01526" target="_blank" rel="noopener"> 266 <i class="fas fa-file-pdf mr-1"></i> Paper 267 </a> 268 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/SlowFast" target="_blank" rel="noopener"> 269 <i class="fas fa-code mr-1"></i> 270 code 271 </a> 272 <br/> 273 274 275 276 </td> 277 </tr> 278 </table> 279 280 281 282 <! ââ MoCo Video ââ> 283 <!-- 284 <table> 285 <tr> 286 <td align="left" width = "50px"></td> 287 <td align="left" width = "350px"> 288 <img class="image" src="image/videomoco.png" width = "250px" alt="" /> 289 <img class="image" src="image/videomoco_output.png" width = "250px" alt="" /> 290 </td> 291 <td align="left" bgcolor="" width = "600px" > 292 <b>A Large-Scale Study on Unsupervised Spatiotemporal Representation Learning</b> <br/> 293 Christoph Feichtenhofer, <
293b>Haoqi Fan</b>, Bo Xiong, Ross Girshick, Kaiming He<br/> 294 <i>Conference on Computer Vision and Pattern Recognition (CVPR)</i>, 2021<br/> 295 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2104.14558" target="_blank" rel="noopener"> 296 <i class="fas fa-file-pdf mr-1"></i> Paper 297 </a> 298 <br/> 299 </td> 300 </tr> 301 </table> 302 --> 303 304 305 <!-- 306 <! ââ Collaborative Mem ââ> 307 <table> 308 <tr> 309 <td align="left" width = "50px"></td> 310 <td align="left" width = "350px"><img class="image" src="image/c_mem.png" width = "250px" alt="" /></td> 311 <td align="left" bgcolor="" width = "600px" > 312 <b>Beyond Short Clips: End-to-End Video-Level Learning with Collaborative Memories</b> <br/> 313 Xitong Yang, <b>Haoqi Fan</b>, Lorenzo Torresani, Larry Davis, Heng Wang <br/> 314 <i>Conference on Computer Vision and Pattern Recognition (CVPR)</i>, 2021<br/> 315 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2104.01198" target="_blank" rel="noopener"> 316 <i class="fas fa-file-pdf mr-1"></i> Paper 317 </a> 318 <br/> 319 </td> 320 </tr> 321 </table> 322 323 324 <! ââ multiview videoââ> 325 <table> 326 <tr> 327 <td align="left" width = "50px"></td> 328 <td align="left" width = "350px"><img class="image" src="image/multiview_video.png" width = "250px" alt="" /></td> 329 <td align="left" bgcolor="" width = "600px" > 330 <b>Multiview Pseudo-Labeling for Semi-supervised Learning from Video</b> <br/> 331 Bo Xiong, <b>Haoqi Fan</b>, Kristen Grauman, Christoph Feichtenhofer<br/> 332 <i>Tech Report</i>, 2021<br/> 333 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2104.00682" target="_blank" rel="noopener"> 334 <i class="fas fa-file-pdf mr-1"></i> Paper 335 </a> 336 <br/> 337 </td> 338 </tr> 339 </table> 340 --> 341 342 343 344 345 346 347 <! ââ MoCo v2 ââ> 348 <!-- 349 <table> 350 <tr> 351 <td align="left" width = "50px"></td> 352 <td align="left" width = "350px"><img class="image" src="image/moco_v2.png" width = "250px" alt="" /></td> 353 <td align="left" bgcolor="" width = "600px" > 354 <b>Improved Baselines with Momentum Contrastive Learning</b> <br/> 355 Xinlei Chen, <b>Haoqi Fan</b>, Ross Girshick, Kaiming He <br/> 356 <i>2-Page Tech Report</i>, 2020<br/> 357 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="http://arxiv.org/abs/2003.04297" target="_blank" rel="noopener"> 358 <i class="fas fa-file-pdf mr-1"></i> Paper 359 </a> 360 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/moco" target="_blank" rel="noopener"> 361 <i class="fas fa-code mr-1"></i> 362 code 363 </a> 364 <br/> 365 </td> 366 </tr> 367 </table> 368 --> 369 370 371 372 <! ââ MoCo ââ> 373 <table> 374 <tr> 375 <td align="left" width = "50px"></td> 376 <td align="left" width = "350px"><img class="image" src="image/moco_v1.png" width = "250px" alt="" /></td> 377 <td align="left" bgcolor="" width = "600px" > 378 <b>Momentum contrast for unsupervised visual representation learning</b> <br/> 379 Kaiming He, <b>Haoqi Fan</b>, Yuxin Wu, Saining Xie, Ross Girshick <br/> 380 <i>Conference on Computer Vision and Pattern Recognition</i> (CVPR), 2020 (Oral)<br/> 381 <i style="color:red">Best Paper Nomination</i><br/> 382 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/1911.05722" target="_blank" rel="noopener"> 383 <i class="fas fa-file-pdf mr-1"></i> Paper 384 </a> 385 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/moco" target="_blank" rel="noopener"> 386 <i class="fas fa-code mr-1"></i> 387 code 388 </a> 389 <br/> 390 <b>Improved Baselines with Momentum Contrastive Learning</b> <br/> 391 Xinlei Chen, <b>Haoqi Fan</b>, Ross Girshick, Kaiming He <br/> 392 <i>2-Page Tech Report</i>, 2020<br/> 393 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="http://arxiv.org/abs/2003.04297" target="_blank" rel="noopener"> 394 <i class="fas fa-file-pdf mr-1"></i> Paper 395 </a> 396 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/moco" target="_blank" rel="noopener"> 397 <i class="fas fa-code mr-1"></i> 398 code 399 </a> 400 <br/> 401 <b>On the Importance of Asymmetry for Siamese Representation Learning</b> <br/> 402 Xiao Wang*, <b>
402Haoqi Fan*</b>, Yuandong Tian, Daisuke Kihara, Xinlei Chen<br/> 403 <i>CVPR</i>, 2022<br/> 404 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="http://arxiv.org/abs/2204.00613" target="_blank" rel="noopener"> 405 <i class="fas fa-file-pdf mr-1"></i> Paper 406 </a> 407 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/asym-siam" target="_blank" rel="noopener"> 408 <i class="fas fa-code mr-1"></i> 409 code 410 </a> 411 <br/> 412 </td> 413 </tr> 414 </table> 415 416 417 <! ââ SlowFast ââ> 418 <table> 419 <tr> 420 <td align="left" width = "50px"></td> 421 <td align="left" width = "350px"><img class="image" src="image/slowfast.gif" width = "250px" alt="" /></td> 422 <td align="left" bgcolor="" width = "600px" > 423 <b>Slowfast networks for video recognition</b> <br/> 424 Christoph Feichtenhofer, <b>Haoqi Fan</b>, Jitendra Malik, Kaiming He <br/> 425 <i>International Conference on Computer Vision</i> (ICCV), 2019 (Oral)<br/> 426 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/1812.03982" target="_blank" rel="noopener"> 427 <i class="fas fa-file-pdf mr-1"></i> Paper 428 </a> 429 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/SlowFast" target="_blank" rel="noopener"> 430 <i class="fas fa-code mr-1"></i> 431 code 432 </a> 433 <br/> 434 </td> 435 </tr> 436 </table> 437 438 439 440 <!-- 441 <! ââ OctConvââ> 442 <table> 443 <tr> 444 <td align="left" width = "50px"></td> 445 <td align="left" width = "350px"><img class="image" src="image/octconv.png" width = "250px" alt="" /></td> 446 <td align="left" bgcolor="" width = "600px" > 447 <b>Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octave Convolution</b> <br/> 448 Yunpeng Chen, <b>Haoqi Fan</b>, Bing Xu, Zhicheng Yan, Yannis Kalantidis, Marcus Rohrbach, Shuicheng Yan, Jiashi Feng <br/> 449 <i>International Conference on Computer Vision</i> (ICCV), 2019 <br/> 450 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/1904.05049" target="_blank" rel="noopener"> 451 <i class="fas fa-file-pdf mr-1"></i> Paper 452 </a> 453 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/OctConv" target="_blank" rel="noopener"> 454 <i class="fas fa-code mr-1"></i> 455 code 456 </a> 457 <br/> 458 </td> 459 </tr> 460 </table> 461 --> 462 463 464 <!-- 465 <! ââ 3D datasetââ> 466 <table> 467 <tr> 468 <td align="left" width = "50px"></td> 469 <td align="left" width = "350px"><img class="image" src="image/3d_dataset.png" width = "250px" alt="" /></td> 470 <td align="left" bgcolor="" width = "600px" > 471 <b>
471Order-Aware Generative Modeling Using the 3D-Craft Dataset</b> <br/> 472 Zhuoyuan Chenâ , Demi Guoâ , Tong Xiaoâ , Saining Xie, Xinlei Chen, Haonan Yu, Jonathan Gray, Kavya Srinet, <b>Haoqi Fan</b>, Jerry Ma, Charles R. Qi, Shubham Tulsiani, Arthur Szlam, and C. Lawrence Zitnick <br/> 473 <i>International Conference on Computer Vision</i> (ICCV), 2019 <br/> 474 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="http://openaccess.thecvf.com/content_ICCV_2019/papers/Chen_Order-Aware_Generative_Modeling_Using_the_3D-Craft_Dataset_ICCV_2019_paper.pdf" target="_blank" rel="noopener"> 475 <i class="fas fa-file-pdf mr-1"></i> Paper 476 </a> 477 <br/> 478 </td> 479 </tr> 480 </table> 481 --> 482 483 484 485 <!-- 486 <! ââ LFB ââ> 487 <table> 488 <tr> 489 <td align="left" width = "50px"></td> 490 <td align="left" width = "350px"><img class="image" src="image/lfb.png" width = "250px" alt="" /></td> 491 <td align="left" bgcolor="" width = "600px" > 492 <b>Long-Term Feature Banks for Detailed Video Understanding</b> <br/> 493 Chao-Yuan Wu, Christoph Feichtenhofer, <b>Haoqi Fan</b>, Kaiming He, Philipp Krähenbühl, Ross Girshick <br/> 494 <i>Conference on Computer Vision and Pattern Recognition</i> (CVPR), 2019 (Oral)<br/> 495 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/1812.05038" target="_blank" rel="noopener"> 496 <i class="fas fa-file-pdf mr-1"></i> Paper 497 </a> 498 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://github.com/facebookresearch/video-long-term-feature-banks" target="_blank" rel="noopener"> 499 <i class="fas fa-code mr-1"></i> 500 code 501 </a> 502 <br/> 503 </td> 504 </tr> 505 </table> 506 --> 507 508 509 <! ââ slaââ> 510 <table> 511 <tr> 512 <td align="left" width = "50px"></td> 513 <td align="left" width = "350px"><img class="image" src="image/sla.png" width = "250px" alt="" /></td> 514 <td align="left" bgcolor="" width = "600px" > 515 <b>Stacked Latent Attention for Multimodal Reasoning</b> <br/> 516 <b>Haoqi Fan</b>, Jiatong Zhou <br/> 517 <i>Conference on Computer Vision and Pattern Recognition</i> (CVPR), 2018 <br/> 518 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="http://openaccess.thecvf.com/content_cvpr_2018/papers/Fan_Stacked_Latent_Attention_CVPR_2018_paper.pdf" target="_blank" rel="noopener"> 519 <i class="fas fa-file-pdf mr-1"></i> Paper 520 </a> 521 <br/> 522 </td> 523 </tr> 524 </table> 525 526 <!-- 527 <! ââ k shotââ> 528 <table> 529 <tr> 530 <td align="left" width = "50px"></td> 531 <td align="left" width = "350px"><img class="image" src="image/k_shot.png" width = "250px" alt="" /></td> 532 <td align="left" bgcolor="" width = "600px" > 533 <b>Efficient K-Shot Learning with Regularized Deep Networks</b> <br/> 534 Donghyun Yoo, <b>Haoqi Fan</b>, Vishnu Naresh Boddeti, Kris M. Kitani <br/> 535 <i>Thirty-Second AAAI Conference on Artificial Intelligence</i> (AAAI), 2018 <br/> 536 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/1710.02277" target="_blank" rel="noopener"> 537 <i class="fas fa-file-pdf mr-1"></i> Paper 538 </a> 539 <br/> 540 </td> 541 </tr> 542 </table> 543 --> 544 545 <! ââ egoââ> 546 <table> 547 <tr> 548 <td align="left" width = "50px"></td> 549 <td align="left" width = "350px"><img class="image" src="image/ego.png" width = "250px" alt="" /></td> 550 <td align="left" bgcolor="" width = "600px" > 551 <b>Going deeper into first-person activity recognition</b> <br/> 552 Minghuang Ma, <b>Haoqi Fan</b>, Kris M. Kitani <br/> 553 <i>Conference on Computer Vision and Pattern Recognition</i> (CVPR), 2016 <br/> 554 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Ma_Going_Deeper_into_CVPR_2016_paper.pdf" target="_blank" rel="noopener"> 555 <i class="fas fa-file-pdf mr-1"></i> Paper 556 </a> 557 <br/> 558 </td> 559 </tr> 560 </table> 561 562 563 <h3>Open Source Projects</h3> 564 565 566 567 <! ââ BAGELââ> 568 <table> 569 <tr> 570 <td align="left" width = "50px"></td> 571 <td align="left" width = "350px"><img class="image" src="image/bagel-promo-vid.gif" width = "250px" alt="" /></td> 572 <td align="left" bgcolor="" width = "600px" > 573 <b>
573BAGEL: The Open-Source Unified Multimodal Model</b> <br/> 574 Chaorui Dengâ, Deyao Zhuâ, Kunchang Liâ, Chenhui Gouâ, Feng Liâ, 575 Zeyu Wang, Shu Zhong, Weihao Yu, Xiaonan Nie, Ziang Song, Guang Shi§, 576 <b>Haoqi Fanââ 577 <a href="https://github.com/bytedance-seed/BAGEL"><img alt="GitHub stars" src="https://img.shields.io/github/stars/bytedance-seed/BAGEL?style=social"></a></li> 578 <br/> 579 </td> 580 </tr> 581 </table> 582 583 584 <! ââ PySlowFastââ> 585 <table> 586 <tr> 587 <td align="left" width = "50px"></td> 588 <td align="left" width = "350px"><img class="image" src="image/pyslowfast.gif" width = "250px" alt="" /></td> 589 <td align="left" bgcolor="" width = "600px" > 590 <b>PySlowFast: video understanding codebase for state-of-the-art research</b> <br/> 591 <b>Haoqi Fan</b>, Yanghao Li, Wan-Yen Lo, Christoph Feichtenhofer<br/> 592 593 <a href="https://github.com/facebookresearch/SlowFast"><img alt="GitHub stars" src="https://img.shields.io/github/stars/facebookresearch/SlowFast?style=social"></a></li> 594 <br/> 595 </td> 596 </tr> 597 </table> 598 599 600 <! ââ PyTorchVideo ââ> 601 <table> 602 <tr> 603 <td align="left" width = "50px"></td> 604 <td align="left" width = "350px"><img style="margin-left:3.5em" class="image" src="image/ptv-teaser.gif" width = "125px" alt="" /></td> 605 <td align="left" bgcolor="" width = "600px" > 606 <b>PyTorchVideo: A Deep Learning Library for Video Understanding</b> <br/> 607 <b>Haoqi Fan *</b>, Tullie Murrell *, <br> Heng Wang <span>‡</span>, Kalyan Vasudev Alwala<span>‡</span>, Yanghao Li<span>‡</span>, Yilei Li<span>‡</span>, Bo Xiong <span>‡</span>, <br> Nikhila Ravi, Meng Li, Haichuan Yang, Jitendra Malik, Ross Girshick, Matt Feiszli, <br> Aaron Adcock<span>†</span>, Wan-Yen Lo<span>†</span>, Christoph Feichtenhofer <span>†</span><br/> 608 609 <a href="https://github.com/facebookresearch/pytorchvideo"><img alt="GitHub stars" src="https://img.shields.io/github/stars/facebookresearch/pytorchvideo?style=social"></a></li> 610 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://arxiv.org/abs/2111.09887?context=cs.LG" target="_blank" rel="noopener"> 611 <i class="fas fa-file-pdf mr-1"></i> Paper 612 </a> 613 <a class="btn btn-outline-primary my-1 mr-1 btn-sm" href="https://ai.facebook.com/blog/pytorchvideo-a-deep-learning-library-for-video-understanding/" target="_blank" rel="noopener"> 614 <i class="fas fa-file-pdf mr-1"></i> Post 615 </a> 616 <br/> 617 </td> 618 </tr> 619 </table> 620 621 622 </div> 623 </div> 624 </div> 625 </div> 626 627 <div id="footer"> 628 629 </div> 630 631 632 633 <div class="container"> 634 <footer class="site-footer"> 635 636 637 <p class="powered-by"> 638 639 © 2016 - 2021 · Haoqi Fan 640 641 642 643 644 <span class="float-right" aria-hidden="true"> 645 <a href="#" id="back_to_top">
646 <span class="button_icon"> 647 <i class="fas fa-chevron-up fa-2x"></i> 648 </span> 649 </a> 650 </span> 651 652 </p> 653</footer> 654 655 </div> 656
657<script>
vendor: 351 bytes, lines 657-663
657 658 (function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ 659 (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), 660 m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) 661 })(window,document,'script','https://www.google-analytics.com/analytics.js','ga'); 662 663 ga('create', '
663UA-88566694-1
vendor: 42 bytes, lines 663-665
663', 'auto'); 664 ga('send', 'pageview'); 665
666</script>
666 667 668
669<script src="js/jquery.min.js"></script>
vendor: 2 bytes, line 669
669
670<script src="js/tether.min.js"></script>
vendor: 2 bytes, line 670
670
671<script src="js/bootstrap.min.js"></script>
vendor: 2 bytes, line 671
671
672<script>
vendor: 351 bytes, lines 672-678
672 673 (function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ 674 (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), 675 m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) 676 })(window,document,'script','https://www.google-analytics.com/analytics.js','ga'); 677 678 ga('create', '
678UA-89924514-1
vendor: 42 bytes, lines 678-680
678', 'auto'); 679 ga('send', 'pageview'); 680
681</script>
681 682
683<script type="text/javascript" src="//rf.revolvermaps.com/0/0/8.js?i=550ynja4lex&m=0c&c=ff8a00&cr1=ff0000&f=arial&l=33&cw=ffffff&cb=007eff" async="async"></script>
683 684 685 686 </body> 687</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.