Deployed 9bcff00c with MkDocs version: 1.0.4

4d567331 · baiyfbupt · 3f842374 · 4d567331 · 4d567331 · 4d567331
隐藏空白更改
内联并排

Showing with 208 addition and 66 deletion

index.html index.html +1 -1

model_zoo/index.html model_zoo/index.html +206 -64

search/search_index.json search/search_index.json +1 -1

sitemap.xml.gz sitemap.xml.gz +0 -0

未找到文件。
--- a/index.html
+++ b/index.html
@@ -293,5 +293,5 @@

 <!--
 MkDocs version : 1.0.4
-Build Date UTC : 2020-01-02 09:26:42
+Build Date UTC : 2020-01-02 11:50:58
 -->
--- a/model_zoo/index.html
+++ b/model_zoo/index.html
@@ -202,7 +202,7 @@
              
                <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>
-<p>数据：ImageNet1000类</p>
+<p>数据集：ImageNet1000类</p>
 <p>评价指标：Top-1/Top-5准确率</p>
 <table>
 <thead>
@@ -215,22 +215,22 @@
 </thead>
 <tbody>
 <tr>
-<td align="center">MobileNetV1</td>
+<td align="center">MobileNetV1 FP32</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="">70.24%/89.03%</a></td>
-<td align="center"><a href="">70.70%/89.55%</a></td>
+<td align="center"><a href="">xx%/xx%</a></td>
+<td align="center"><a href="">xx%/xx%</a></td>
 </tr>
 <tr>
-<td align="center">MobileNetV2</td>
+<td align="center">MobileNetV2 FP32</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="">71.36%/90.17%</a></td>
-<td align="center"><a href="">72.02%/90.23%</a></td>
+<td align="center"></td>
+<td align="center"></td>
 </tr>
 <tr>
-<td align="center">ResNet50</td>
+<td align="center">ResNet50 FP32</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="">76.26%/92.81%</a></td>
-<td align="center"><a href="">76.59%/93.04%</a></td>
+<td align="center"></td>
+<td align="center"></td>
 </tr>
 </tbody>
 </table>
@@ -248,12 +248,12 @@
 <tr>
 <td align="left">MobileNetV1</td>
 <td align="center">17M</td>
-<td align="center"></td>
-<td align="center"></td>
+<td align="center"><a href="">xx%/xx%</a></td>
+<td align="center"><a href="">xx%/xx%</a></td>
 </tr>
 <tr>
 <td align="left">MobileNetV2</td>
-<td align="center"></td>
+<td align="center">xxM</td>
 <td align="center"></td>
 <td align="center"></td>
 </tr>
@@ -266,40 +266,77 @@
 </tbody>
 </table>
 <h3 id="12">1.2 目标检测<a class="headerlink" href="#12" title="Permanent link">#</a></h3>
-<p>数据：COCO 2017 </p>
-<p>评价指标：mAP</p>
-<p>输入尺寸：608</p>
+<p>数据集：COCO 2017 </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>
+<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>
 </tr>
 </thead>
 <tbody>
 <tr>
 <td align="center">MobileNet-V1-YOLOv3</td>
+<td align="center">608</td>
+<td align="center">8</td>
 <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="">27.9</a></td>
-<td align="center"><a href="">28.0</a></td>
+<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>
 </tr>
 <tr>
 <td align="center">R50-dcn-YOLOv3</td>
+<td align="center">608</td>
+<td align="center"></td>
 <td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_pretrained_coco.tar">41.4</a></td>
-<td align="center"><a href="">40.4</a></td>
-<td align="center"><a href="">40.6</a></td>
+<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>
 </tr>
 </tbody>
 </table>
-<p>数据：WIDER-FACE</p>
+<p>数据集：WIDER-FACE</p>
 <p>评价指标：Easy/Medium/Hard mAP</p>
-<p>输入尺寸：640</p>
 <table>
 <thead>
 <tr>
 <th align="center">Model</th>
+<th align="center">输入尺寸</th>
+<th align="center">Image/GPU</th>
 <th align="center">FP32</th>
 <th align="center">离线量化</th>
 <th align="center">量化训练</th>
@@ -308,18 +345,24 @@
 <tbody>
 <tr>
 <td align="center">BlazeFace</td>
+<td align="center">640</td>
+<td align="center">8</td>
 <td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/blazeface_original.tar">0.915/0.892/0.797</a></td>
-<td align="center"></td>
-<td align="center"></td>
+<td align="center"><a href="">xx/xx/xx</a></td>
+<td align="center"><a href="">xx/xx/xx</a></td>
 </tr>
 <tr>
 <td align="center">BlazeFace-Lite</td>
+<td align="center">640</td>
+<td align="center"></td>
 <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>
+<td align="center">640</td>
+<td align="center"></td>
 <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>
@@ -327,23 +370,22 @@
 </tbody>
 </table>
 <h3 id="13">1.3 图像分割<a class="headerlink" href="#13" title="Permanent link">#</a></h3>
-<p>数据：Cityscapes</p>
-<p>评价指标：mIoU</p>
+<p>数据集：Cityscapes</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>
+<th align="center">FP32 mIoU</th>
+<th align="center">离线量化 mIoU</th>
+<th align="center">量化训练 mIoU</th>
 </tr>
 </thead>
 <tbody>
 <tr>
 <td align="center">DeepLabv3+/MobileNetv1</td>
 <td align="center"><a href="">63.26</a></td>
-<td align="center"></td>
-<td align="center"></td>
+<td align="center"><a href="">xx</a></td>
+<td align="center"><a href="">xx</a></td>
 </tr>
 <tr>
 <td align="center">DeepLabv3+/MobileNetv2</td>
@@ -356,91 +398,155 @@
 <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>
-<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">均匀剪枝</th>
-<th align="center">敏感度剪枝</th>
-<th align="center">自动剪枝</th>
+<th align="center">Top-1/Top-5</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="">69.4%/88.66%</a></td>
-<td align="center"><a href="">69.8%/88.9%</a></td>
-<td align="center">-</td>
+</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>
 </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="">65.79%/86.11%</a></td>
-<td align="center">-</td>
-<td align="center">-</td>
+</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>
 </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>
-<td align="center"><a href="">70.99%/89.95%</a></td>
-<td align="center">-</td>
-<td align="center"><a href="">70.24%/89.63%</a></td>
+</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>
 </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>
-<p>评价指标：mAP</p>
-<p>输入尺寸：608</p>
 <table>
 <thead>
 <tr>
 <th align="center">Model</th>
 <th align="center">数据集</th>
-<th align="center">baseline</th>
-<th align="center">敏感度剪枝</th>
+<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>
 </tr>
 </thead>
 <tbody>
 <tr>
 <td align="center">MobileNet-V1-YOLOv3</td>
 <td align="center">Pasacl VOC</td>
+<td align="center">608</td>
+<td align="center">8</td>
 <td align="center"><a href="">76.2</a></td>
-<td align="center"><a href="">77.59</a></td>
+<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>
 </tr>
 <tr>
 <td align="center">MobileNet-V1-YOLOv3</td>
 <td align="center">COCO</td>
+<td align="center">608</td>
+<td align="center"></td>
 <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="">29.56</a></td>
+<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>
 </tr>
 <tr>
 <td align="center">R50-dcn-YOLOv3</td>
 <td align="center">COCO</td>
+<td align="center">608</td>
+<td align="center"></td>
 <td align="center"><a href="https://paddlemodels.bj.bcebos.com/object_detection/yolov3_r50vd_dcn_obj365_pretrained_coco.tar">41.4</a></td>
-<td align="center"><a href="">37.8</a></td>
+<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>
 </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>
-<p>评价指标：mIoU</p>
 <table>
 <thead>
 <tr>
 <th align="center">Model</th>
-<th align="center">Baseline</th>
-<th align="center">剪枝</th>
+<th align="center">Baseline mIoU</th>
+<th align="center">xx剪枝 mIoU</th>
 </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>
-<td align="center"></td>
+<td align="center"><a href="">xx</a></td>
 </tr>
 </tbody>
 </table>
@@ -476,36 +582,72 @@
 </table>
 <div class="admonition note">
 <p class="admonition-title">Note</p>
-<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后缀代表是开启了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>
+<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>
 <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>
-<p>评价指标：mAP</p>
-<p>输入尺寸：608</p>
 <table>
 <thead>
 <tr>
 <th align="center">Model</th>
 <th align="center">数据集</th>
+<th align="center">输入尺寸</th>
+<th align="center">Image/GPU</th>
 <th align="center">baseline</th>
-<th align="center">蒸馏后</th>
+<th align="center">蒸馏后 mAP</th>
 </tr>
 </thead>
 <tbody>
 <tr>
 <td align="center">MobileNet-V1-YOLOv3</td>
 <td align="center">Pasacl VOC</td>
+<td align="center">640</td>
+<td align="center">16</td>
 <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>
+<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>
 <td align="center">COCO</td>
+<td align="center">640</td>
+<td align="center"></td>
 <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>
+<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>
 </tbody>
 </table>
 <div class="admonition note">

--- a/search/search_index.json
+++ b/search/search_index.json
--- a/sitemap.xml.gz
+++ b/sitemap.xml.gz