pytorch主流backbone实现

灰太狼 2023-10-02 10:35 21阅读 0赞

计算机视觉主流backbone收集

源码 [https://github.com/oaifaye/backbone-pytorch][https_github.com_oaifaye_backbone-pytorch]

<table> 
 <thead> 
  <tr> 
   <th>dataset</th> 
   <th>network</th> 
   <th>params</th> 
   <th>top1 err</th> 
   <th>top5 err</th> 
   <th>epoch(lr = 0.1)</th> 
   <th>epoch(lr = 0.02)</th> 
   <th>epoch(lr = 0.004)</th> 
   <th>epoch(lr = 0.0008)</th> 
   <th>total epoch</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>cifar100</td> 
   <td>mobilenet</td> 
   <td>3.3M</td> 
   <td>34.02</td> 
   <td>10.56</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>mobilenetv2</td> 
   <td>2.36M</td> 
   <td>31.92</td> 
   <td>09.02</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>squeezenet</td> 
   <td>0.78M</td> 
   <td>30.59</td> 
   <td>8.36</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>shufflenet</td> 
   <td>1.0M</td> 
   <td>29.94</td> 
   <td>8.35</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>shufflenetv2</td> 
   <td>1.3M</td> 
   <td>30.49</td> 
   <td>8.49</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>vgg11_bn</td> 
   <td>28.5M</td> 
   <td>31.36</td> 
   <td>11.85</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>vgg13_bn</td> 
   <td>28.7M</td> 
   <td>28.00</td> 
   <td>9.71</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>vgg16_bn</td> 
   <td>34.0M</td> 
   <td>27.07</td> 
   <td>8.84</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>vgg19_bn</td> 
   <td>39.0M</td> 
   <td>27.77</td> 
   <td>8.84</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>resnet18</td> 
   <td>11.2M</td> 
   <td>24.39</td> 
   <td>6.95</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>resnet34</td> 
   <td>21.3M</td> 
   <td>23.24</td> 
   <td>6.63</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>resnet50</td> 
   <td>23.7M</td> 
   <td>22.61</td> 
   <td>6.04</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>resnet101</td> 
   <td>42.7M</td> 
   <td>22.22</td> 
   <td>5.61</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>resnet152</td> 
   <td>58.3M</td> 
   <td>22.31</td> 
   <td>5.81</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>preactresnet18</td> 
   <td>11.3M</td> 
   <td>27.08</td> 
   <td>8.53</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>preactresnet34</td> 
   <td>21.5M</td> 
   <td>24.79</td> 
   <td>7.68</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>preactresnet50</td> 
   <td>23.9M</td> 
   <td>25.73</td> 
   <td>8.15</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>preactresnet101</td> 
   <td>42.9M</td> 
   <td>24.84</td> 
   <td>7.83</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>preactresnet152</td> 
   <td>58.6M</td> 
   <td>22.71</td> 
   <td>6.62</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>resnext50</td> 
   <td>14.8M</td> 
   <td>22.23</td> 
   <td>6.00</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>resnext101</td> 
   <td>25.3M</td> 
   <td>22.22</td> 
   <td>5.99</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>resnext152</td> 
   <td>33.3M</td> 
   <td>22.40</td> 
   <td>5.58</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>attention59</td> 
   <td>55.7M</td> 
   <td>33.75</td> 
   <td>12.90</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>attention92</td> 
   <td>102.5M</td> 
   <td>36.52</td> 
   <td>11.47</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>densenet121</td> 
   <td>7.0M</td> 
   <td>22.99</td> 
   <td>6.45</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>densenet161</td> 
   <td>26M</td> 
   <td>21.56</td> 
   <td>6.04</td> 
   <td>60</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>densenet201</td> 
   <td>18M</td> 
   <td>21.46</td> 
   <td>5.9</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>googlenet</td> 
   <td>6.2M</td> 
   <td>21.97</td> 
   <td>5.94</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>inceptionv3</td> 
   <td>22.3M</td> 
   <td>22.81</td> 
   <td>6.39</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>inceptionv4</td> 
   <td>41.3M</td> 
   <td>24.14</td> 
   <td>6.90</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>inceptionresnetv2</td> 
   <td>65.4M</td> 
   <td>27.51</td> 
   <td>9.11</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>xception</td> 
   <td>21.0M</td> 
   <td>25.07</td> 
   <td>7.32</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>seresnet18</td> 
   <td>11.4M</td> 
   <td>23.56</td> 
   <td>6.68</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>seresnet34</td> 
   <td>21.6M</td> 
   <td>22.07</td> 
   <td>6.12</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>seresnet50</td> 
   <td>26.5M</td> 
   <td>21.42</td> 
   <td>5.58</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>seresnet101</td> 
   <td>47.7M</td> 
   <td>20.98</td> 
   <td>5.41</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>seresnet152</td> 
   <td>66.2M</td> 
   <td>20.66</td> 
   <td>5.19</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>nasnet</td> 
   <td>5.2M</td> 
   <td>22.71</td> 
   <td>5.91</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>wideresnet-40-10</td> 
   <td>55.9M</td> 
   <td>21.25</td> 
   <td>5.77</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>stochasticdepth18</td> 
   <td>11.22M</td> 
   <td>31.40</td> 
   <td>8.84</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>stochasticdepth34</td> 
   <td>21.36M</td> 
   <td>27.72</td> 
   <td>7.32</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>stochasticdepth50</td> 
   <td>23.71M</td> 
   <td>23.35</td> 
   <td>5.76</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
  <tr> 
   <td>cifar100</td> 
   <td>stochasticdepth101</td> 
   <td>42.69M</td> 
   <td>21.28</td> 
   <td>5.39</td> 
   <td>60</td> 
   <td>60</td> 
   <td>40</td> 
   <td>40</td> 
   <td>200</td> 
  </tr> 
 </tbody> 
</table>

[https_github.com_oaifaye_backbone-pytorch]: https://github.com/oaifaye/backbone-pytorch