Skip to content

P
PaddleSlim

PaddlePaddle / PaddleSlim
1 年多 前同步成功

通知 51
Star 1434
Fork 344
P
PaddleSlim
切换分支/标签
查找文件 普通视图历史永久链接Permalink
index.html 21.0 KB
Newer Older
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
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 
<!DOCTYPE html>
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
  <meta charset="utf-8">
  <meta http-equiv="X-UA-Compatible" content="IE=edge">
  <meta name="viewport" content="width=device-width, initial-scale=1.0">
  
  
  <link rel="shortcut icon" href="../img/favicon.ico">
  <title>模型库 - PaddleSlim Docs</title>
  <link href='https://fonts.googleapis.com/css?family=Lato:400,700|Roboto+Slab:400,700|Inconsolata:400,700' rel='stylesheet' type='text/css'>

  <link rel="stylesheet" href="../css/theme.css" type="text/css" />
  <link rel="stylesheet" href="../css/theme_extra.css" type="text/css" />
  <link rel="stylesheet" href="//cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css">
  
  <script>
    // Current page data
    var mkdocs_page_name = "\u6a21\u578b\u5e93";
    var mkdocs_page_input_path = "model_zoo.md";
    var mkdocs_page_url = null;
  </script>
  
  <script src="../js/jquery-2.1.1.min.js" defer></script>
  <script src="../js/modernizr-2.8.3.min.js" defer></script>
  <script src="//cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
  <script>hljs.initHighlightingOnLoad();</script> 
  
</head>

<body class="wy-body-for-nav" role="document">

  <div class="wy-grid-for-nav">

    
    <nav data-toggle="wy-nav-shift" class="wy-nav-side stickynav">
      <div class="wy-side-nav-search">
        <a href=".." class="icon icon-home"> PaddleSlim Docs</a>
        <div role="search">
  <form id ="rtd-search-form" class="wy-form" action="../search.html" method="get">
    <input type="text" name="q" placeholder="Search docs" title="Type search term here" />
  </form>
</div>
      </div>

      <div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation">
	<ul class="current">
	  
          
            <li class="toctree-l1">
		
    <a class="" href="..">Home</a>
	    </li>
          
            <li class="toctree-l1">
		
    <span class="caption-text">API</span>
    <ul class="subnav">
                <li class="">
                    
    <a class="" href="../api/quantization_api/">量化</a>
                </li>
                <li class="">
                    
    <a class="" href="../api/prune_api/">剪枝与敏感度</a>
                </li>
                <li class="">
                    
    <a class="" href="../api/analysis_api/">模型分析</a>
                </li>
                <li class="">
                    
    <a class="" href="../api/single_distiller_api/">知识蒸馏</a>
                </li>
                <li class="">
                    
    <a class="" href="../api/nas_api/">SA搜索</a>
                </li>
                <li class="">
                    
    <a class="" href="../api/search_space/">搜索空间</a>
                </li>
                <li class="">
                    
    <a class="" href="../table_latency/">硬件延时评估表</a>
                </li>
    </ul>
	    </li>
          
            <li class="toctree-l1">
		
    <span class="caption-text">教程</span>
    <ul class="subnav">
                <li class="">
                    
    <a class="" href="../tutorials/quant_post_demo/">离线量化</a>
                </li>
                <li class="">
                    
    <a class="" href="../tutorials/quant_aware_demo/">量化训练</a>
                </li>
                <li class="">
                    
    <a class="" href="../tutorials/quant_embedding_demo/">Embedding量化</a>
                </li>
                <li class="">
                    
    <a class="" href="../tutorials/nas_demo/">SA搜索</a>
                </li>
                <li class="">
                    
    <a class="" href="../tutorials/distillation_demo/">知识蒸馏</a>
                </li>
    </ul>
	    </li>
          
            <li class="toctree-l1">
		
    <a class="" href="../algo/algo/">算法原理</a>
	    </li>
          
            <li class="toctree-l1 current">
		
    <a class="current" href="./">模型库</a>
    <ul class="subnav">
            
    <li class="toctree-l2"><a href="#1">1. 量化</a></li>
    
        <ul>
        
            <li><a class="toctree-l3" href="#11">1.1 图象分类</a></li>
        
            <li><a class="toctree-l3" href="#12">1.2 目标检测</a></li>
        
            <li><a class="toctree-l3" href="#13">1.3 图像分割</a></li>
        
        </ul>
    

    <li class="toctree-l2"><a href="#2">2. 剪枝</a></li>
    
        <ul>
        
            <li><a class="toctree-l3" href="#21">2.1 图像分类</a></li>
        
            <li><a class="toctree-l3" href="#22">2.2 目标检测</a></li>
        
            <li><a class="toctree-l3" href="#23">2.3 图像分割</a></li>
        
        </ul>
    

    <li class="toctree-l2"><a href="#3">3. 蒸馏</a></li>
    
        <ul>
        
            <li><a class="toctree-l3" href="#31">3.1 图象分类</a></li>
        
            <li><a class="toctree-l3" href="#32">3.2 目标检测</a></li>
        
        </ul>
    

    </ul>
	    </li>
          
        </ul>
      </div>
      &nbsp;
    </nav>

    <section data-toggle="wy-nav-shift" class="wy-nav-content-wrap">

      
      <nav class="wy-nav-top" role="navigation" aria-label="top navigation">
        <i data-toggle="wy-nav-top" class="fa fa-bars"></i>
        <a href="..">PaddleSlim Docs</a>
      </nav>

      
      <div class="wy-nav-content">
        <div class="rst-content">
          <div role="navigation" aria-label="breadcrumbs navigation">
  <ul class="wy-breadcrumbs">
    <li><a href="..">Docs</a> &raquo;</li>
    
      
    
    <li>模型库</li>
    <li class="wy-breadcrumbs-aside">
      
        <a href="https://github.com/PaddlePaddle/PaddleSlim/edit/master/docs/model_zoo.md"
          class="icon icon-github"> Edit on GitHub</a>
      
    </li>
  </ul>
  <hr/>
</div>
          <div role="main">
            <div class="section">
              
                <h2 id="1">1. 量化<a class="headerlink" href="#1" title="Permanent link">#</a></h2>
<h3 id="11">1.1 图象分类<a class="headerlink" href="#11" title="Permanent link">#</a></h3>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
205 
<p>数据集：ImageNet1000类</p>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
206 207 208 209 210 211 212 213 214 215 216 217 
<p>评价指标：Top-1/Top-5准确率</p>
<table>
<thead>
<tr>
<th align="center">Model</th>
<th align="center">FP32</th>
<th align="center">离线量化</th>
<th align="center">量化训练</th>
</tr>
</thead>
<tbody>
<tr>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
218 
<td align="center">MobileNetV1 FP32</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
219 
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV1_pretrained.tar">70.99%/89.68%</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
220 221 
<td align="center"><a href="">xx%/xx%</a></td>
<td align="center"><a href="">xx%/xx%</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
222 223 
</tr>
<tr>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
224 
<td align="center">MobileNetV2 FP32</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
225 
<td align="center"><a href="https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_pretrained.tar">72.15%/90.65%</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
226 227 
<td align="center"></td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
228 229 
</tr>
<tr>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
230 
<td align="center">ResNet50 FP32</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
231 
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_pretrained.tar">76.50%/93.00%</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
232 233 
<td align="center"></td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 
</tr>
</tbody>
</table>
<p>量化训练前后，模型大小的变化对比如下：</p>
<table>
<thead>
<tr>
<th align="left">Model</th>
<th align="center">FP32</th>
<th align="center">离线量化</th>
<th align="center">量化训练</th>
</tr>
</thead>
<tbody>
<tr>
<td align="left">MobileNetV1</td>
<td align="center">17M</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
251 252 
<td align="center"><a href="">xx%/xx%</a></td>
<td align="center"><a href="">xx%/xx%</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
253 254 255 
</tr>
<tr>
<td align="left">MobileNetV2</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
256 
<td align="center">xxM</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
257 258 259 260 261 262 263 264 265 266 267 268 
<td align="center"></td>
<td align="center"></td>
</tr>
<tr>
<td align="left">ResNet50</td>
<td align="center">99M</td>
<td align="center"></td>
<td align="center"></td>
</tr>
</tbody>
</table>
<h3 id="12">1.2 目标检测<a class="headerlink" href="#12" title="Permanent link">#</a></h3>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
269 
<p>数据集：COCO 2017 </p>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
270 271 272 273 
<table>
<thead>
<tr>
<th align="center">Model</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
274 275 276 277 278 
<th align="center">输入尺寸</th>
<th align="center">Image/GPU</th>
<th align="center">FP32 BoxAP</th>
<th align="center">离线量化 BoxAP</th>
<th align="center">量化训练 BoxAP</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
279 280 281 282 283 
</tr>
</thead>
<tbody>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
284 285 
<td align="center">608</td>
<td align="center">8</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
286 
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar">29.3</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 
<td align="center"><a href="">xx</a></td>
<td align="center"><a href="">xx</a></td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">416</td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">320</td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
305 306 307 
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
308 309 
<td align="center">608</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
310 
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_pretrained_coco.tar">41.4</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 
<td align="center"></td>
<td align="center"></td>
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3</td>
<td align="center">416</td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3</td>
<td align="center">320</td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
329 330 331 
</tr>
</tbody>
</table>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
332 
<p>数据集：WIDER-FACE</p>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
333 334 335 336 337 
<p>评价指标：Easy/Medium/Hard mAP</p>
<table>
<thead>
<tr>
<th align="center">Model</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
338 339 
<th align="center">输入尺寸</th>
<th align="center">Image/GPU</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
340 341 342 343 344 345 346 347 
<th align="center">FP32</th>
<th align="center">离线量化</th>
<th align="center">量化训练</th>
</tr>
</thead>
<tbody>
<tr>
<td align="center">BlazeFace</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
348 349 
<td align="center">640</td>
<td align="center">8</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
350 
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/blazeface_original.tar">0.915/0.892/0.797</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
351 352 
<td align="center"><a href="">xx/xx/xx</a></td>
<td align="center"><a href="">xx/xx/xx</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
353 354 355 
</tr>
<tr>
<td align="center">BlazeFace-Lite</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
356 357 
<td align="center">640</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
358 359 360 361 362 363 
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/blazeface_lite.tar">0.909/0.885/0.781</a></td>
<td align="center"></td>
<td align="center"></td>
</tr>
<tr>
<td align="center">BlazeFace-NAS</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
364 365 
<td align="center">640</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
366 367 368 369 370 371 372 
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/blazeface_nas.tar">0.837/0.807/0.658</a></td>
<td align="center"></td>
<td align="center"></td>
</tr>
</tbody>
</table>
<h3 id="13">1.3 图像分割<a class="headerlink" href="#13" title="Permanent link">#</a></h3>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
373 
<p>数据集：Cityscapes</p>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
374 375 376 377 
<table>
<thead>
<tr>
<th align="center">Model</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
378 379 380 
<th align="center">FP32 mIoU</th>
<th align="center">离线量化 mIoU</th>
<th align="center">量化训练 mIoU</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
381 382 383 384 385 386 
</tr>
</thead>
<tbody>
<tr>
<td align="center">DeepLabv3+/MobileNetv1</td>
<td align="center"><a href="">63.26</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
387 388 
<td align="center"><a href="">xx</a></td>
<td align="center"><a href="">xx</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 
</tr>
<tr>
<td align="center">DeepLabv3+/MobileNetv2</td>
<td align="center"><a href="https://paddleseg.bj.bcebos.com/models/mobilenet_cityscapes.tgz">69.81</a></td>
<td align="center"></td>
<td align="center"></td>
</tr>
</tbody>
</table>
<h2 id="2">2. 剪枝<a class="headerlink" href="#2" title="Permanent link">#</a></h2>
<h3 id="21">2.1 图像分类<a class="headerlink" href="#21" title="Permanent link">#</a></h3>
<p>数据：ImageNet1000类</p>
<table>
<thead>
<tr>
<th align="center">Model</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
405 
<th align="center">Top-1/Top-5</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
406 407 408 409 410 411 
</tr>
</thead>
<tbody>
<tr>
<td align="center">MobileNetV1</td>
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV1_pretrained.tar">70.99%/89.68%</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
412 413 414 415 416 417 418 419 
</tr>
<tr>
<td align="center">MobileNetV1 uniform -50%</td>
<td align="center"></td>
</tr>
<tr>
<td align="center">MobileNetV1 sensitive -xx%</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
420 421 422 423 
</tr>
<tr>
<td align="center">MobileNetV2</td>
<td align="center"><a href="https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_pretrained.tar">72.15%/90.65%</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
424 425 426 427 428 429 430 431 
</tr>
<tr>
<td align="center">MobileNetV2 uniform -50%</td>
<td align="center"></td>
</tr>
<tr>
<td align="center">MobileNetV2 sensitive -xx%</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
432 433 434 435 
</tr>
<tr>
<td align="center">ResNet34</td>
<td align="center"><a href="https://paddle-imagenet-models-name.bj.bcebos.com/ResNet34_pretrained.tar">74.57%/92.14%</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
436 437 438 439 440 441 442 443 
</tr>
<tr>
<td align="center">ResNet34 uniform -50%</td>
<td align="center"></td>
</tr>
<tr>
<td align="center">ResNet34 auto -50%</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
444 445 446 447 448 449 450 451 452 453 
</tr>
</tbody>
</table>
<h3 id="22">2.2 目标检测<a class="headerlink" href="#22" title="Permanent link">#</a></h3>
<p>数据：Pasacl VOC &amp; COCO 2017</p>
<table>
<thead>
<tr>
<th align="center">Model</th>
<th align="center">数据集</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
454 455 456 457 
<th align="center">输入尺寸</th>
<th align="center">Image/GPU</th>
<th align="center">baseline mAP</th>
<th align="center">敏感度剪枝 mAP</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
458 459 460 461 462 463 
</tr>
</thead>
<tbody>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">Pasacl VOC</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
464 465 
<td align="center">608</td>
<td align="center">8</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
466 
<td align="center"><a href="">76.2</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 
<td align="center"><a href="">77.59</a> (-50%)</td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">Pasacl VOC</td>
<td align="center">416</td>
<td align="center"></td>
<td align="center"><a href="">76.7</a></td>
<td align="center"><a href="">xx</a></td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">Pasacl VOC</td>
<td align="center">320</td>
<td align="center"></td>
<td align="center"><a href="">75.2</a></td>
<td align="center"><a href="">xx</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
484 485 486 487 
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">COCO</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
488 489 
<td align="center">608</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
490 
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar">29.3</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 
<td align="center"><a href="">29.56</a> (-20%)</td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">COCO</td>
<td align="center">416</td>
<td align="center"></td>
<td align="center"><a href="">29.3</a></td>
<td align="center"><a href="">xx</a></td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">COCO</td>
<td align="center">320</td>
<td align="center"></td>
<td align="center"><a href="">27.1</a></td>
<td align="center"><a href="">xx</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
508 509 510 511 
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3</td>
<td align="center">COCO</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
512 513 
<td align="center">608</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
514 
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_pretrained_coco.tar">41.4</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 
<td align="center"><a href="">37.8</a> (-30%)</td>
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3</td>
<td align="center">COCO</td>
<td align="center">416</td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3</td>
<td align="center">COCO</td>
<td align="center">320</td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
532 533 534 535 536 537 538 539 540 
</tr>
</tbody>
</table>
<h3 id="23">2.3 图像分割<a class="headerlink" href="#23" title="Permanent link">#</a></h3>
<p>数据：Cityscapes</p>
<table>
<thead>
<tr>
<th align="center">Model</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
541 542 
<th align="center">Baseline mIoU</th>
<th align="center">xx剪枝 mIoU</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
543 544 545 546 547 548 
</tr>
</thead>
<tbody>
<tr>
<td align="center">DeepLabv3+/MobileNetv2</td>
<td align="center"><a href="https://paddleseg.bj.bcebos.com/models/mobilenet_cityscapes.tgz">69.81</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
549 
<td align="center"><a href="">xx</a></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 
</tr>
</tbody>
</table>
<h2 id="3">3. 蒸馏<a class="headerlink" href="#3" title="Permanent link">#</a></h2>
<h3 id="31">3.1 图象分类<a class="headerlink" href="#31" title="Permanent link">#</a></h3>
<p>数据：ImageNet1000类</p>
<p>评价指标：Top-1/Top-5准确率</p>
<table>
<thead>
<tr>
<th align="center">Model</th>
<th align="center">baseline</th>
<th align="center">蒸馏后</th>
</tr>
</thead>
<tbody>
<tr>
<td align="center">MobileNetV1</td>
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV1_pretrained.tar">70.99%/89.68%</a></td>
<td align="center"><a href="">72.79%/90.69%</a> (teacher: ResNet50_vd<sup><a href="#trans1">1</a></sup>)</td>
</tr>
<tr>
<td align="center">MobileNetV2</td>
<td align="center"><a href="https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_pretrained.tar">72.15%/90.65%</a></td>
<td align="center"><a href="">74.30%/91.52%</a> (teacher: ResNet50_vd)</td>
</tr>
<tr>
<td align="center">ResNet50</td>
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_pretrained.tar">76.50%/93.00%</a></td>
<td align="center"><a href="">77.40%/93.48%</a> (teacher: ResNet101<sup><a href="#trans2">2</a></sup>)</td>
</tr>
</tbody>
</table>
<div class="admonition note">
<p class="admonition-title">Note</p>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
585 586 
<p><a name="trans1">  [1]</a>：<a href="https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_pretrained.tar">ResNet50_vd</a>预训练模型Top-1/Top-5准确率分别为79.12%/94.44%</p>
<p>带_vd后缀代表开启了Mixup训练，Mixup相关介绍参考<a href="https://arxiv.org/abs/1710.09412">mixup: Beyond Empirical Risk Minimization</a></p>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
587 588 589 590 591 592 593 594 595 
<p><a name="trans1">[2]</a>：<a href="https://paddle-imagenet-models-name.bj.bcebos.com/ResNet101_pretrained.tar">ResNet101</a>预训练模型Top-1/Top-5准确率分别为77.56%/93.64%</p>
</div>
<h3 id="32">3.2 目标检测<a class="headerlink" href="#32" title="Permanent link">#</a></h3>
<p>数据：Pasacl VOC &amp; COCO 2017</p>
<table>
<thead>
<tr>
<th align="center">Model</th>
<th align="center">数据集</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
596 597 
<th align="center">输入尺寸</th>
<th align="center">Image/GPU</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
598 
<th align="center">baseline</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
599 
<th align="center">蒸馏后 mAP</th>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
600 601 602 603 604 605 
</tr>
</thead>
<tbody>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">Pasacl VOC</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
606 607 
<td align="center">640</td>
<td align="center">16</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
608 609 610 611 612 
<td align="center"><a href="">76.2</a></td>
<td align="center"><a href="">79.0</a> (teacher: ResNet34-YOLOv3-VOC<sup><a href="#trans3">3</a></sup>)</td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 
<td align="center">Pasacl VOC</td>
<td align="center">416</td>
<td align="center"></td>
<td align="center"><a href="">76.7</a></td>
<td align="center"><a href="">78.2</a></td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">Pasacl VOC</td>
<td align="center">320</td>
<td align="center"></td>
<td align="center"><a href="">75.2</a></td>
<td align="center"><a href="">75.5</a></td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
629 
<td align="center">COCO</td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
630 631 
<td align="center">640</td>
<td align="center"></td>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
632 633 634 
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar">29.3</a></td>
<td align="center"><a href="">31.0</a> (teacher: ResNet34-YOLOv3-COCO<sup><a href="#trans4">4</a></sup>)</td>
</tr>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">COCO</td>
<td align="center">416</td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">COCO</td>
<td align="center">320</td>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
</tr>
B
Deployed 9bcff00 with MkDocs version: 1.0.4  
baiyfbupt 已提交
651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 
</tbody>
</table>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p><a name="trans1">[3]</a>：<a href="">ResNet34-YOLOv3-VOC</a>预训练模型的Box AP为82.6</p>
<p><a name="trans1">[4]</a>：<a href="">ResNet34-YOLOv3-COCO</a>预训练模型的Box AP为36.2</p>
</div>
              
            </div>
          </div>
          <footer>
  
    <div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
      
      
        <a href="../algo/algo/" class="btn btn-neutral" title="算法原理"><span class="icon icon-circle-arrow-left"></span> Previous</a>
      
    </div>
  

  <hr/>

  <div role="contentinfo">
    <!-- Copyright etc -->
    
  </div>

  Built with <a href="http://www.mkdocs.org">MkDocs</a> using a <a href="https://github.com/snide/sphinx_rtd_theme">theme</a> provided by <a href="https://readthedocs.org">Read the Docs</a>.
</footer>
      
        </div>
      </div>

    </section>

  </div>

  <div class="rst-versions" role="note" style="cursor: pointer">
    <span class="rst-current-version" data-toggle="rst-current-version">
      
          <a href="https://github.com/PaddlePaddle/PaddleSlim/" class="fa fa-github" style="float: left; color: #fcfcfc"> GitHub</a>
      
      
        <span><a href="../algo/algo/" style="color: #fcfcfc;">&laquo; Previous</a></span>
      
      
    </span>
</div>
    <script>var base_url = '..';</script>
    <script src="../js/theme.js" defer></script>
      <script src="../mathjax-config.js" defer></script>
      <script src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0/MathJax.js?config=TeX-AMS-MML_HTMLorMML" defer></script>
      <script src="../search/main.js" defer></script>

</body>
</html>