index.html 31.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
<!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">
17
  <link href="../extra.css" rel="stylesheet">
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
  
  <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>
          
57 58 59 60 61
            <li class="toctree-l1 current">
		
    <a class="current" href="./">模型库</a>
    <ul class="subnav">
            
62
    <li class="toctree-l2"><a href="#1">1. 图象分类</a></li>
63 64 65 66 67
    
        <ul>
        
            <li><a class="toctree-l3" href="#11">1.1 量化</a></li>
        
68
            <li><a class="toctree-l3" href="#12">1.2 剪裁</a></li>
69 70 71 72 73 74 75 76 77 78 79 80
        
            <li><a class="toctree-l3" href="#13">1.3 蒸馏</a></li>
        
        </ul>
    

    <li class="toctree-l2"><a href="#2">2. 目标检测</a></li>
    
        <ul>
        
            <li><a class="toctree-l3" href="#21">2.1 量化</a></li>
        
81
            <li><a class="toctree-l3" href="#22">2.2 剪裁</a></li>
82 83 84 85 86 87 88 89 90 91 92 93
        
            <li><a class="toctree-l3" href="#23">2.3 蒸馏</a></li>
        
        </ul>
    

    <li class="toctree-l2"><a href="#3">3. 图像分割</a></li>
    
        <ul>
        
            <li><a class="toctree-l3" href="#31">3.1 量化</a></li>
        
94
            <li><a class="toctree-l3" href="#32">3.2 剪裁</a></li>
95 96 97 98 99 100 101
        
        </ul>
    

    </ul>
	    </li>
          
102 103
            <li class="toctree-l1">
		
104
    <span class="caption-text">教程</span>
105 106 107
    <ul class="subnav">
                <li class="">
                    
108
    <a class="" href="../tutorials/quant_post_demo/">离线量化</a>
109 110 111
                </li>
                <li class="">
                    
112
    <a class="" href="../tutorials/quant_aware_demo/">量化训练</a>
113 114 115
                </li>
                <li class="">
                    
116
    <a class="" href="../tutorials/quant_embedding_demo/">Embedding量化</a>
117 118 119
                </li>
                <li class="">
                    
120
    <a class="" href="../tutorials/nas_demo/">SA搜索</a>
121 122 123
                </li>
                <li class="">
                    
124
    <a class="" href="../search_space/">搜索空间</a>
125 126 127
                </li>
                <li class="">
                    
128
    <a class="" href="../tutorials/distillation_demo/">知识蒸馏</a>
129 130 131 132 133 134
                </li>
    </ul>
	    </li>
          
            <li class="toctree-l1">
		
135
    <span class="caption-text">API</span>
136 137 138
    <ul class="subnav">
                <li class="">
                    
139
    <a class="" href="../api/quantization_api/">量化</a>
140 141 142
                </li>
                <li class="">
                    
143
    <a class="" href="../api/prune_api/">剪枝与敏感度</a>
144 145 146
                </li>
                <li class="">
                    
147
    <a class="" href="../api/analysis_api/">模型分析</a>
148 149 150
                </li>
                <li class="">
                    
151
    <a class="" href="../api/single_distiller_api/">知识蒸馏</a>
152 153 154
                </li>
                <li class="">
                    
155 156 157 158 159
    <a class="" href="../api/nas_api/">SA搜索</a>
                </li>
                <li class="">
                    
    <a class="" href="../table_latency/">硬件延时评估表</a>
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
                </li>
    </ul>
	    </li>
          
            <li class="toctree-l1">
		
    <a class="" href="../algo/algo/">算法原理</a>
	    </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">
      
194
        <a href="https://github.com/PaddlePaddle/PaddleSlim/edit/master/docs/model_zoo.md"
195 196 197 198 199 200 201 202 203
          class="icon icon-github"> Edit on GitHub</a>
      
    </li>
  </ul>
  <hr/>
</div>
          <div role="main">
            <div class="section">
              
204
                <h2 id="1">1. 图象分类<a class="headerlink" href="#1" title="Permanent link">#</a></h2>
205
<p>数据集:ImageNet1000类</p>
206
<h3 id="11">1.1 量化<a class="headerlink" href="#11" title="Permanent link">#</a></h3>
207 208 209
<table>
<thead>
<tr>
210 211 212
<th align="center">模型</th>
<th align="center">压缩方法</th>
<th align="center">Top-1/Top-5 Acc</th>
213
<th align="center">模型体积(MB)</th>
214
<th align="center">下载</th>
215 216 217 218
</tr>
</thead>
<tbody>
<tr>
219
<td align="center">MobileNetV1</td>
220
<td align="center">-</td>
221 222 223
<td align="center">70.99%/89.68%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
224 225
</tr>
<tr>
226 227 228 229 230
<td align="center">MobileNetV1</td>
<td align="center">quant_post</td>
<td align="center">xx%/xx%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
231 232
</tr>
<tr>
233 234 235 236 237
<td align="center">MobileNetV1</td>
<td align="center">quant_aware</td>
<td align="center">xx%/xx%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
238 239
</tr>
<tr>
240
<td align="center">MobileNetV2</td>
241
<td align="center">-</td>
242 243 244
<td align="center">72.15%/90.65%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
245 246
</tr>
<tr>
247 248 249 250 251
<td align="center">MobileNetV2</td>
<td align="center">quant_post</td>
<td align="center">xx%/xx%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
252 253
</tr>
<tr>
254
<td align="center">MobileNetV2</td>
255 256 257 258 259 260 261
<td align="center">quant_aware</td>
<td align="center">xx%/xx%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
</tr>
<tr>
<td align="center">ResNet50</td>
262
<td align="center">-</td>
263 264 265
<td align="center">76.50%/93.00%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
266 267
</tr>
<tr>
268
<td align="center">ResNet50</td>
269 270 271 272 273 274 275 276 277 278 279
<td align="center">quant_post</td>
<td align="center">xx%/xx%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
</tr>
<tr>
<td align="center">ResNet50</td>
<td align="center">quant_aware</td>
<td align="center">xx%/xx%</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
280 281 282
</tr>
</tbody>
</table>
283
<h3 id="12">1.2 剪裁<a class="headerlink" href="#12" title="Permanent link">#</a></h3>
284 285 286
<table>
<thead>
<tr>
287 288 289
<th align="center">模型</th>
<th align="center">压缩方法</th>
<th align="center">Top-1/Top-5 Acc</th>
290
<th align="center">模型体积(MB)</th>
291
<th align="center">GFLOPs</th>
292
<th align="center">下载</th>
293 294 295 296
</tr>
</thead>
<tbody>
<tr>
297
<td align="center">MobileNetV1</td>
298
<td align="center">Baseline</td>
299
<td align="center">70.99%/89.68%</td>
300 301
<td align="center">17</td>
<td align="center">1.11</td>
302
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV1_pretrained.tar">下载链接</a></td>
303 304
</tr>
<tr>
305
<td align="center">MobileNetV1</td>
306
<td align="center">uniform -50%</td>
307
<td align="center">69.4%/88.66% (-1.59%/-1.02%)</td>
308 309
<td align="center">9</td>
<td align="center">0.56</td>
310
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/MobileNetV1_uniform-50.tar">下载链接</a></td>
311 312
</tr>
<tr>
313
<td align="center">MobileNetV1</td>
314
<td align="center">sensitive -30%</td>
315
<td align="center">70.4%/89.3% (-0.59%/-0.38%)</td>
316 317
<td align="center">12</td>
<td align="center">0.74</td>
318
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/MobileNetV1_sensitive-30.tar">下载链接</a></td>
319 320
</tr>
<tr>
321 322
<td align="center">MobileNetV1</td>
<td align="center">sensitive -50%</td>
323
<td align="center">69.8% / 88.9% (-1.19%/-0.78%)</td>
324 325
<td align="center">9</td>
<td align="center">0.56</td>
326
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/MobileNetV1_sensitive-50.tar">下载链接</a></td>
327 328
</tr>
<tr>
329
<td align="center">MobileNetV2</td>
330
<td align="center">-</td>
331 332 333
<td align="center">72.15%/90.65%</td>
<td align="center">15</td>
<td align="center">0.59</td>
334
<td align="center"><a href="https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_pretrained.tar">下载链接</a></td>
335 336
</tr>
<tr>
337
<td align="center">MobileNetV2</td>
338
<td align="center">uniform -50%</td>
339
<td align="center">65.79%/86.11% (-6.35%/-4.47%)</td>
340 341
<td align="center">11</td>
<td align="center">0.296</td>
342
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/MobileNetV2_uniform-50.tar">下载链接</a></td>
343
</tr>
344
<tr>
345
<td align="center">ResNet34</td>
346
<td align="center">-</td>
347 348 349
<td align="center">72.15%/90.65%</td>
<td align="center">84</td>
<td align="center">7.36</td>
350
<td align="center"><a href="https://paddle-imagenet-models-name.bj.bcebos.com/ResNet34_pretrained.tar">下载链接</a></td>
351 352
</tr>
<tr>
353
<td align="center">ResNet34</td>
354
<td align="center">uniform -50%</td>
355
<td align="center">70.99%/89.95% (-1.36%/-0.87%)</td>
356 357
<td align="center">41</td>
<td align="center">3.67</td>
358
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/ResNet34_uniform-50.tar">下载链接</a></td>
359 360
</tr>
<tr>
361
<td align="center">ResNet34</td>
362
<td align="center">auto -55.05%</td>
363
<td align="center">70.24%/89.63% (-2.04%/-1.06%)</td>
364 365
<td align="center">33</td>
<td align="center">3.31</td>
366
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/ResNet34_auto-55.tar">下载链接</a></td>
367 368 369
</tr>
</tbody>
</table>
370
<h3 id="13">1.3 蒸馏<a class="headerlink" href="#13" title="Permanent link">#</a></h3>
371 372 373
<table>
<thead>
<tr>
374 375 376
<th align="center">模型</th>
<th align="center">压缩方法</th>
<th align="center">Top-1/Top-5 Acc</th>
377
<th align="center">模型体积(MB)</th>
378
<th align="center">下载</th>
379 380 381 382
</tr>
</thead>
<tbody>
<tr>
383
<td align="center">MobileNetV1</td>
384
<td align="center">student</td>
385 386
<td align="center">70.99%/89.68%</td>
<td align="center">17</td>
387
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV1_pretrained.tar">下载链接</a></td>
388 389
</tr>
<tr>
390
<td align="center">ResNet50_vd</td>
391
<td align="center">teacher</td>
392 393
<td align="center">79.12%/94.44%</td>
<td align="center">99</td>
394
<td align="center"><a href="https://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_vd_pretrained.tar">下载链接</a></td>
395 396
</tr>
<tr>
397 398
<td align="center">MobileNetV1</td>
<td align="center">ResNet50_vd<sup><a href="#trans1">1</a></sup> distill</td>
399
<td align="center">72.77%/90.68% (+1.78%/+1.00%)</td>
400
<td align="center">17</td>
401
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/MobileNetV1_distilled.tar">下载链接</a></td>
402
</tr>
403
<tr>
404
<td align="center">MobileNetV2</td>
405
<td align="center">student</td>
406 407
<td align="center">72.15%/90.65%</td>
<td align="center">15</td>
408
<td align="center"><a href="https://paddle-imagenet-models-name.bj.bcebos.com/MobileNetV2_pretrained.tar">下载链接</a></td>
409 410
</tr>
<tr>
411 412
<td align="center">MobileNetV2</td>
<td align="center">ResNet50_vd distill</td>
413
<td align="center">74.28%/91.53% (+2.13%/+0.88%)</td>
414
<td align="center">15</td>
415
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/MobileNetV2_distilled.tar">下载链接</a></td>
416 417
</tr>
<tr>
418
<td align="center">ResNet50</td>
419
<td align="center">student</td>
420 421
<td align="center">76.50%/93.00%</td>
<td align="center">99</td>
422
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/ResNet50_pretrained.tar">下载链接</a></td>
423 424
</tr>
<tr>
425
<td align="center">ResNet101</td>
426
<td align="center">teacher</td>
427 428
<td align="center">77.56%/93.64%</td>
<td align="center">173</td>
429
<td align="center"><a href="http://paddle-imagenet-models-name.bj.bcebos.com/ResNet101_pretrained.tar">下载链接</a></td>
430 431 432 433
</tr>
<tr>
<td align="center">ResNet50</td>
<td align="center">ResNet101 distill</td>
434
<td align="center">77.29%/93.65% (+0.79%/+0.65%)</td>
435
<td align="center">99</td>
436
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/ResNet50_distilled.tar">下载链接</a></td>
437 438 439
</tr>
</tbody>
</table>
440 441 442 443 444 445 446
<div class="admonition note">
<p class="admonition-title">Note</p>
<p><a name="trans1">[1]</a>:带_vd后缀代表该预训练模型使用了Mixup,Mixup相关介绍参考<a href="https://arxiv.org/abs/1710.09412">mixup: Beyond Empirical Risk Minimization</a></p>
</div>
<h2 id="2">2. 目标检测<a class="headerlink" href="#2" title="Permanent link">#</a></h2>
<h3 id="21">2.1 量化<a class="headerlink" href="#21" title="Permanent link">#</a></h3>
<p>数据集: COCO 2017</p>
447 448 449
<table>
<thead>
<tr>
450 451 452 453 454 455 456
<th align="center">模型</th>
<th align="center">压缩方法</th>
<th align="center">数据集</th>
<th align="center">Image/GPU</th>
<th align="center">输入608 Box AP</th>
<th align="center">输入416 Box AP</th>
<th align="center">输入320 Box AP</th>
457
<th align="center">模型体积(MB)</th>
458
<th align="center">下载</th>
459 460 461 462
</tr>
</thead>
<tbody>
<tr>
463
<td align="center">MobileNet-V1-YOLOv3</td>
464
<td align="center">-</td>
465 466 467 468 469 470 471
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">29.3</td>
<td align="center">29.3</td>
<td align="center">27.1</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
472 473
</tr>
<tr>
474 475 476 477 478 479 480 481 482
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">quant_post</td>
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
483
</tr>
484
<tr>
485 486 487 488 489 490 491 492 493
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">quant_aware</td>
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
494 495
</tr>
<tr>
496
<td align="center">R50-dcn-YOLOv3 obj365_pretrain</td>
497
<td align="center">-</td>
498 499 500 501 502 503 504
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">41.4</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
505 506
</tr>
<tr>
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526
<td align="center">R50-dcn-YOLOv3 obj365_pretrain</td>
<td align="center">quant_post</td>
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3 obj365_pretrain</td>
<td align="center">quant_aware</td>
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
527 528 529
</tr>
</tbody>
</table>
530
<p>数据集:WIDER-FACE</p>
531 532 533
<table>
<thead>
<tr>
534 535 536 537 538
<th align="center">模型</th>
<th align="center">压缩方法</th>
<th align="center">Image/GPU</th>
<th align="center">输入尺寸</th>
<th align="center">Easy/Medium/Hard</th>
539
<th align="center">模型体积(MB)</th>
540
<th align="center">下载</th>
541 542 543 544
</tr>
</thead>
<tbody>
<tr>
545
<td align="center">BlazeFace</td>
546
<td align="center">-</td>
547 548 549 550 551
<td align="center">8</td>
<td align="center">640</td>
<td align="center">0.915/0.892/0.797</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
552 553
</tr>
<tr>
554 555 556 557 558 559 560
<td align="center">BlazeFace</td>
<td align="center">quant_post</td>
<td align="center">8</td>
<td align="center">640</td>
<td align="center">xx/xx/xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
561 562
</tr>
<tr>
563 564 565 566 567 568 569
<td align="center">BlazeFace</td>
<td align="center">quant_aware</td>
<td align="center">8</td>
<td align="center">640</td>
<td align="center">xx/xx/xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
570 571
</tr>
<tr>
572
<td align="center">BlazeFace-Lite</td>
573
<td align="center">-</td>
574 575 576 577 578
<td align="center">8</td>
<td align="center">640</td>
<td align="center">0.909/0.885/0.781</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
579 580
</tr>
<tr>
581 582 583 584 585 586 587
<td align="center">BlazeFace-Lite</td>
<td align="center">quant_post</td>
<td align="center">8</td>
<td align="center">640</td>
<td align="center">xx/xx/xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
588
</tr>
589
<tr>
590 591 592 593 594 595 596
<td align="center">BlazeFace-Lite</td>
<td align="center">quant_aware</td>
<td align="center">8</td>
<td align="center">640</td>
<td align="center">xx/xx/xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
597 598
</tr>
<tr>
599
<td align="center">BlazeFace-NAS</td>
600
<td align="center">-</td>
601 602 603
<td align="center">8</td>
<td align="center">640</td>
<td align="center">0.837/0.807/0.658</td>
604
<td align="center">xx</td>
605
<td align="center"><a href="">下载链接</a></td>
606 607
</tr>
<tr>
608 609 610 611 612
<td align="center">BlazeFace-NAS</td>
<td align="center">quant_post</td>
<td align="center">8</td>
<td align="center">640</td>
<td align="center">xx/xx/xx</td>
613
<td align="center">xx</td>
614
<td align="center"><a href="">下载链接</a></td>
615 616
</tr>
<tr>
617 618 619 620 621
<td align="center">BlazeFace-NAS</td>
<td align="center">quant_aware</td>
<td align="center">8</td>
<td align="center">640</td>
<td align="center">xx/xx/xx</td>
622
<td align="center">xx</td>
623
<td align="center"><a href="">下载链接</a></td>
624 625 626
</tr>
</tbody>
</table>
627
<h3 id="22">2.2 剪裁<a class="headerlink" href="#22" title="Permanent link">#</a></h3>
628
<p>数据集:Pasacl VOC &amp; COCO 2017</p>
629 630 631
<table>
<thead>
<tr>
632 633
<th align="center">模型</th>
<th align="center">压缩方法</th>
634
<th align="center">数据集</th>
635
<th align="center">Image/GPU</th>
636 637 638
<th align="center">输入608 Box AP</th>
<th align="center">输入416 Box AP</th>
<th align="center">输入320 Box AP</th>
639
<th align="center">模型体积(MB)</th>
640
<th align="center">GFLOPs (608*608)</th>
641
<th align="center">下载</th>
642 643 644 645 646
</tr>
</thead>
<tbody>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
647
<td align="center">Baseline</td>
648
<td align="center">Pascal VOC</td>
649
<td align="center">8</td>
650 651 652
<td align="center">76.2</td>
<td align="center">76.7</td>
<td align="center">75.3</td>
653 654 655
<td align="center">94</td>
<td align="center">40.49</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar">下载链接</a></td>
656 657 658
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
659 660
<td align="center">sensitive -52.88%</td>
<td align="center">Pascal VOC</td>
661
<td align="center">8</td>
662 663 664
<td align="center">77.6 (+1.4)</td>
<td align="center">77.7 (1.0)</td>
<td align="center">75.5 (+0.2)</td>
665 666 667
<td align="center">31</td>
<td align="center">19.08</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/yolov3_mobilenet_v1_voc_prune.tar">下载链接</a></td>
668 669 670
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
671
<td align="center">-</td>
672 673 674 675
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">29.3</td>
<td align="center">29.3</td>
676 677 678 679
<td align="center">27.0</td>
<td align="center">95</td>
<td align="center">41.35</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar">下载链接</a></td>
680 681 682
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
683
<td align="center">sensitive -51.77%</td>
684
<td align="center">COCO</td>
685
<td align="center">8</td>
686 687 688
<td align="center">26.0 (-3.3)</td>
<td align="center">25.1 (-4.2)</td>
<td align="center">22.6 (-4.4)</td>
689 690 691
<td align="center">32</td>
<td align="center">19.94</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/yolov3_mobilenet_v1_prune.tar">下载链接</a></td>
692 693
</tr>
<tr>
694
<td align="center">R50-dcn-YOLOv3</td>
695
<td align="center">-</td>
696
<td align="center">COCO</td>
697 698
<td align="center">8</td>
<td align="center">39.1</td>
699 700 701 702 703
<td align="center">-</td>
<td align="center">-</td>
<td align="center">177</td>
<td align="center">89.60</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn.tar">下载链接</a></td>
704 705
</tr>
<tr>
706
<td align="center">R50-dcn-YOLOv3</td>
707
<td align="center">sensitive -9.37%</td>
708
<td align="center">COCO</td>
709
<td align="center">8</td>
710
<td align="center">39.3 (+0.2)</td>
711 712 713 714 715
<td align="center">-</td>
<td align="center">-</td>
<td align="center">150</td>
<td align="center">81.20</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/yolov3_r50vd_dcn_prune.tar">下载链接</a></td>
716 717 718
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3</td>
719
<td align="center">sensitive -24.68%</td>
720
<td align="center">COCO</td>
721
<td align="center">8</td>
722
<td align="center">37.3 (-1.8)</td>
723 724 725 726 727
<td align="center">-</td>
<td align="center">-</td>
<td align="center">113</td>
<td align="center">67.48</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/yolov3_r50vd_dcn_prune578.tar">下载链接</a></td>
728 729
</tr>
<tr>
730
<td align="center">R50-dcn-YOLOv3 obj365_pretrain</td>
731
<td align="center">-</td>
732 733 734
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">41.4</td>
735 736 737 738 739
<td align="center">-</td>
<td align="center">-</td>
<td align="center">177</td>
<td align="center">89.60</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_pretrained_coco.tar">下载链接</a></td>
740 741 742
</tr>
<tr>
<td align="center">R50-dcn-YOLOv3 obj365_pretrain</td>
743
<td align="center">sensitive -9.37%</td>
744
<td align="center">COCO</td>
745
<td align="center">8</td>
746
<td align="center">40.5 (-0.9)</td>
747 748 749 750 751
<td align="center">-</td>
<td align="center">-</td>
<td align="center">150</td>
<td align="center">81.20</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/yolov3_r50vd_dcn_obj365_pretrained_coco_prune.tar">下载链接</a></td>
752 753
</tr>
<tr>
754
<td align="center">R50-dcn-YOLOv3 obj365_pretrain</td>
755
<td align="center">sensitive -24.68%</td>
756
<td align="center">COCO</td>
757
<td align="center">8</td>
758
<td align="center">37.8 (-3.3)</td>
759 760 761 762 763
<td align="center">-</td>
<td align="center">-</td>
<td align="center">113</td>
<td align="center">67.48</td>
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/yolov3_r50vd_dcn_obj365_pretrained_coco_prune578.tar">下载链接</a></td>
764 765 766
</tr>
</tbody>
</table>
767 768
<h3 id="23">2.3 蒸馏<a class="headerlink" href="#23" title="Permanent link">#</a></h3>
<p>数据集:Pasacl VOC &amp; COCO 2017</p>
769 770 771
<table>
<thead>
<tr>
772 773 774 775 776 777 778
<th align="center">模型</th>
<th align="center">压缩方法</th>
<th align="center">数据集</th>
<th align="center">Image/GPU</th>
<th align="center">输入608 Box AP</th>
<th align="center">输入416 Box AP</th>
<th align="center">输入320 Box AP</th>
779
<th align="center">模型体积(MB)</th>
780
<th align="center">下载</th>
781 782 783 784
</tr>
</thead>
<tbody>
<tr>
785
<td align="center">MobileNet-V1-YOLOv3</td>
786
<td align="center">-</td>
787 788 789 790 791 792
<td align="center">Pascal VOC</td>
<td align="center">8</td>
<td align="center">76.2</td>
<td align="center">76.7</td>
<td align="center">75.3</td>
<td align="center">94</td>
793
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1_voc.tar">下载链接</a></td>
794 795
</tr>
<tr>
796
<td align="center">ResNet34-YOLOv3</td>
797
<td align="center">-</td>
798 799 800 801 802 803
<td align="center">Pascal VOC</td>
<td align="center">8</td>
<td align="center">82.6</td>
<td align="center">81.9</td>
<td align="center">80.1</td>
<td align="center">162</td>
804
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34_voc.tar">下载链接</a></td>
805 806
</tr>
<tr>
807 808 809 810
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">ResNet34-YOLOv3 distill</td>
<td align="center">Pascal VOC</td>
<td align="center">8</td>
811 812 813
<td align="center">79.0 (+2.8)</td>
<td align="center">78.2 (+1.5)</td>
<td align="center">75.5 (+0.2)</td>
814
<td align="center">94</td>
815
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/yolov3_mobilenetv1_voc_distilled.tar">下载链接</a></td>
816
</tr>
817
<tr>
818
<td align="center">MobileNet-V1-YOLOv3</td>
819
<td align="center">-</td>
820 821 822 823 824 825
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">29.3</td>
<td align="center">29.3</td>
<td align="center">27.0</td>
<td align="center">95</td>
826
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_mobilenet_v1.tar">下载链接</a></td>
827 828
</tr>
<tr>
829
<td align="center">ResNet34-YOLOv3</td>
830
<td align="center">-</td>
831 832 833 834 835 836
<td align="center">COCO</td>
<td align="center">8</td>
<td align="center">36.2</td>
<td align="center">34.3</td>
<td align="center">31.4</td>
<td align="center">163</td>
837
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r34.tar">下载链接</a></td>
838 839 840 841 842 843
</tr>
<tr>
<td align="center">MobileNet-V1-YOLOv3</td>
<td align="center">ResNet34-YOLOv3 distill</td>
<td align="center">COCO</td>
<td align="center">8</td>
844 845 846
<td align="center">31.4 (+2.1)</td>
<td align="center">30.0 (+0.7)</td>
<td align="center">27.1 (+0.1)</td>
847
<td align="center">95</td>
848
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/yolov3_mobilenetv1_coco_distilled.tar">下载链接</a></td>
849 850 851
</tr>
</tbody>
</table>
852 853 854
<h2 id="3">3. 图像分割<a class="headerlink" href="#3" title="Permanent link">#</a></h2>
<p>数据集:Cityscapes</p>
<h3 id="31">3.1 量化<a class="headerlink" href="#31" title="Permanent link">#</a></h3>
855 856 857
<table>
<thead>
<tr>
858 859 860
<th align="center">模型</th>
<th align="center">压缩方法</th>
<th align="center">mIoU</th>
861
<th align="center">模型体积(MB)</th>
862
<th align="center">下载</th>
863 864 865 866
</tr>
</thead>
<tbody>
<tr>
867
<td align="center">DeepLabv3+/MobileNetv1</td>
868
<td align="center">-</td>
869 870 871 872 873 874 875
<td align="center">63.26</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
</tr>
<tr>
<td align="center">DeepLabv3+/MobileNetv1</td>
<td align="center">quant_post</td>
876 877
<td align="center">xx</td>
<td align="center">xx</td>
878
<td align="center"><a href="">下载链接</a></td>
879
</tr>
880
<tr>
881 882 883 884 885
<td align="center">DeepLabv3+/MobileNetv1</td>
<td align="center">quant_aware</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
886 887
</tr>
<tr>
888
<td align="center">DeepLabv3+/MobileNetv2</td>
889
<td align="center">-</td>
890 891 892
<td align="center">69.81</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
893 894
</tr>
<tr>
895 896 897 898 899
<td align="center">DeepLabv3+/MobileNetv2</td>
<td align="center">quant_post</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
900 901
</tr>
<tr>
902 903 904 905 906
<td align="center">DeepLabv3+/MobileNetv2</td>
<td align="center">quant_aware</td>
<td align="center">xx</td>
<td align="center">xx</td>
<td align="center"><a href="">下载链接</a></td>
907 908 909
</tr>
</tbody>
</table>
910
<h3 id="32">3.2 剪裁<a class="headerlink" href="#32" title="Permanent link">#</a></h3>
911 912 913
<table>
<thead>
<tr>
914 915 916
<th align="center">模型</th>
<th align="center">压缩方法</th>
<th align="center">mIoU</th>
917
<th align="center">模型体积(MB)</th>
918
<th align="center">GFLOPs</th>
919
<th align="center">下载</th>
920 921 922 923
</tr>
</thead>
<tbody>
<tr>
924 925 926 927 928
<td align="center">fast-scnn</td>
<td align="center">baseline</td>
<td align="center">69.64</td>
<td align="center">11</td>
<td align="center">14.41</td>
929
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/fast_scnn_cityscape.tar">下载链接</a></td>
930 931
</tr>
<tr>
932 933 934 935 936
<td align="center">fast-scnn</td>
<td align="center">uniform  -17.07%</td>
<td align="center">69.58 (-0.06)</td>
<td align="center">8.5</td>
<td align="center">11.95</td>
937
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/fast_scnn_cityscape_uniform-17.tar">下载链接</a></td>
938 939 940 941 942 943 944
</tr>
<tr>
<td align="center">fast-scnn</td>
<td align="center">sensitive -47.60%</td>
<td align="center">66.68 (-2.96)</td>
<td align="center">5.7</td>
<td align="center">7.55</td>
945
<td align="center"><a href="https://paddlemodels.bj.bcebos.com/PaddleSlim/fast_scnn_cityscape_sensitive-47.tar">下载链接</a></td>
946
</tr>
947 948 949 950 951 952 953 954 955
</tbody>
</table>
              
            </div>
          </div>
          <footer>
  
    <div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
      
956
        <a href="../tutorials/quant_post_demo/" class="btn btn-neutral float-right" title="离线量化">Next <span class="icon icon-circle-arrow-right"></span></a>
957
      
958
      
959
        <a href=".." class="btn btn-neutral" title="Home"><span class="icon icon-circle-arrow-left"></span> Previous</a>
960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986
      
    </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>
      
      
987
        <span><a href=".." style="color: #fcfcfc;">&laquo; Previous</a></span>
988 989
      
      
990
        <span style="margin-left: 15px"><a href="../tutorials/quant_post_demo/" style="color: #fcfcfc">Next &raquo;</a></span>
991
      
992 993 994 995 996 997 998 999 1000 1001
    </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>