rhetenor/awesome-awesomeness
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
862b240c90a9bccce579388e075561631b8fbc77
awesome-awesomeness/html/computervisionmodels.html
Jonas Zeunert 2d63fe63cd Restructure with Makefile + add html output
2024-04-20 19:22:54 +02:00

1972 lines
76 KiB
HTML
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
<h1 id="awesome-computer-vision-models-awesome">Awesome Computer Vision
Models <a href="https://awesome.re"><img
src="https://awesome.re/badge-flat.svg" alt="Awesome" /></a></h1>
<p>A curated list of popular classification, segmentation and detection
models with corresponding evaluation metrics from papers.</p>
<h2 id="contents">Contents</h2>
<ul>
<li><a href="#classification-models">Classification models</a></li>
<li><a href="#segmentation-models">Segmentation models</a></li>
<li><a href="#detection-models">Detection models</a></li>
</ul>
<h2 id="classification-models">Classification models</h2>
<table>
<colgroup>
<col style="width: 75%" />
<col style="width: 6%" />
<col style="width: 5%" />
<col style="width: 5%" />
<col style="width: 4%" />
<col style="width: 2%" />
</colgroup>
<thead>
<tr class="header">
<th style="text-align: center;">Model</th>
<th style="text-align: center;">Number of parameters</th>
<th style="text-align: center;">FLOPS</th>
<th style="text-align: center;">Top-1 Error</th>
<th style="text-align: center;">Top-5 Error</th>
<th style="text-align: center;">Year</th>
</tr>
</thead>
<tbody>
<tr class="odd">
<td style="text-align: center;">AlexNet (<a
href="https://arxiv.org/abs/1404.5997">‘One weird trick for
parallelizing convolutional neural networks’</a>)</td>
<td style="text-align: center;">62.3M</td>
<td style="text-align: center;">1,132.33M</td>
<td style="text-align: center;">40.96</td>
<td style="text-align: center;">18.24</td>
<td style="text-align: center;">2014</td>
</tr>
<tr class="even">
<td style="text-align: center;">VGG-16 (<a
href="https://arxiv.org/abs/1409.1556">‘Very Deep Convolutional Networks
for Large-Scale Image Recognition’</a>)</td>
<td style="text-align: center;">138.3M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">26.78</td>
<td style="text-align: center;">8.69</td>
<td style="text-align: center;">2014</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ResNet-10 (<a
href="https://arxiv.org/abs/1512.03385">‘Deep Residual Learning for
Image Recognition’</a>)</td>
<td style="text-align: center;">5.5M</td>
<td style="text-align: center;">894.04M</td>
<td style="text-align: center;">34.69</td>
<td style="text-align: center;">14.36</td>
<td style="text-align: center;">2015</td>
</tr>
<tr class="even">
<td style="text-align: center;">ResNet-18 (<a
href="https://arxiv.org/abs/1512.03385">‘Deep Residual Learning for
Image Recognition’</a>)</td>
<td style="text-align: center;">11.7M</td>
<td style="text-align: center;">1,820.41M</td>
<td style="text-align: center;">28.53</td>
<td style="text-align: center;">9.82</td>
<td style="text-align: center;">2015</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ResNet-34 (<a
href="https://arxiv.org/abs/1512.03385">‘Deep Residual Learning for
Image Recognition’</a>)</td>
<td style="text-align: center;">21.8M</td>
<td style="text-align: center;">3,672.68M</td>
<td style="text-align: center;">24.84</td>
<td style="text-align: center;">7.80</td>
<td style="text-align: center;">2015</td>
</tr>
<tr class="even">
<td style="text-align: center;">ResNet-50 (<a
href="https://arxiv.org/abs/1512.03385">‘Deep Residual Learning for
Image Recognition’</a>)</td>
<td style="text-align: center;">25.5M</td>
<td style="text-align: center;">3,877.95M</td>
<td style="text-align: center;">22.28</td>
<td style="text-align: center;">6.33</td>
<td style="text-align: center;">2015</td>
</tr>
<tr class="odd">
<td style="text-align: center;">InceptionV3 (<a
href="https://arxiv.org/abs/1512.00567">‘Rethinking the Inception
Architecture for Computer Vision’</a>)</td>
<td style="text-align: center;">23.8M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">21.2</td>
<td style="text-align: center;">5.6</td>
<td style="text-align: center;">2015</td>
</tr>
<tr class="even">
<td style="text-align: center;">PreResNet-18 (<a
href="https://arxiv.org/abs/1603.05027">‘Identity Mappings in Deep
Residual Networks’</a>)</td>
<td style="text-align: center;">11.7M</td>
<td style="text-align: center;">1,820.56M</td>
<td style="text-align: center;">28.43</td>
<td style="text-align: center;">9.72</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">PreResNet-34 (<a
href="https://arxiv.org/abs/1603.05027">‘Identity Mappings in Deep
Residual Networks’</a>)</td>
<td style="text-align: center;">21.8M</td>
<td style="text-align: center;">3,672.83M</td>
<td style="text-align: center;">24.89</td>
<td style="text-align: center;">7.74</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">PreResNet-50 (<a
href="https://arxiv.org/abs/1603.05027">‘Identity Mappings in Deep
Residual Networks’</a>)</td>
<td style="text-align: center;">25.6M</td>
<td style="text-align: center;">3,875.44M</td>
<td style="text-align: center;">22.40</td>
<td style="text-align: center;">6.47</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DenseNet-121 (<a
href="https://arxiv.org/abs/1608.06993">‘Densely Connected Convolutional
Networks’</a>)</td>
<td style="text-align: center;">8.0M</td>
<td style="text-align: center;">2,872.13M</td>
<td style="text-align: center;">23.48</td>
<td style="text-align: center;">7.04</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">DenseNet-161 (<a
href="https://arxiv.org/abs/1608.06993">‘Densely Connected Convolutional
Networks’</a>)</td>
<td style="text-align: center;">28.7M</td>
<td style="text-align: center;">7,793.16M</td>
<td style="text-align: center;">22.86</td>
<td style="text-align: center;">6.44</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">PyramidNet-101 (<a
href="https://arxiv.org/abs/1610.02915">‘Deep Pyramidal Residual
Networks’</a>)</td>
<td style="text-align: center;">42.5M</td>
<td style="text-align: center;">8,743.54M</td>
<td style="text-align: center;">21.98</td>
<td style="text-align: center;">6.20</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">ResNeXt-14(32x4d) (<a
href="http://arxiv.org/abs/1611.05431">‘Aggregated Residual
Transformations for Deep Neural Networks’</a>)</td>
<td style="text-align: center;">9.5M</td>
<td style="text-align: center;">1,603.46M</td>
<td style="text-align: center;">30.32</td>
<td style="text-align: center;">11.46</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ResNeXt-26(32x4d) (<a
href="http://arxiv.org/abs/1611.05431">‘Aggregated Residual
Transformations for Deep Neural Networks’</a>)</td>
<td style="text-align: center;">15.4M</td>
<td style="text-align: center;">2,488.07M</td>
<td style="text-align: center;">24.14</td>
<td style="text-align: center;">7.46</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">WRN-50-2 (<a
href="https://arxiv.org/abs/1605.07146">‘Wide Residual
Networks’</a>)</td>
<td style="text-align: center;">68.9M</td>
<td style="text-align: center;">11,405.42M</td>
<td style="text-align: center;">22.53</td>
<td style="text-align: center;">6.41</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">Xception (<a
href="https://arxiv.org/abs/1610.02357">‘Xception: Deep Learning with
Depthwise Separable Convolutions’</a>)</td>
<td style="text-align: center;">22,855,952</td>
<td style="text-align: center;">8,403.63M</td>
<td style="text-align: center;">20.97</td>
<td style="text-align: center;">5.49</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">InceptionV4 (<a
href="https://arxiv.org/abs/1602.07261">‘Inception-v4, Inception-ResNet
and the Impact of Residual Connections on Learning’</a>)</td>
<td style="text-align: center;">42,679,816</td>
<td style="text-align: center;">12,304.93M</td>
<td style="text-align: center;">20.64</td>
<td style="text-align: center;">5.29</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">InceptionResNetV2 (<a
href="https://arxiv.org/abs/1602.07261">‘Inception-v4, Inception-ResNet
and the Impact of Residual Connections on Learning’</a>)</td>
<td style="text-align: center;">55,843,464</td>
<td style="text-align: center;">13,188.64M</td>
<td style="text-align: center;">19.93</td>
<td style="text-align: center;">4.90</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">PolyNet (<a
href="https://arxiv.org/abs/1611.05725">‘PolyNet: A Pursuit of
Structural Diversity in Very Deep Networks’</a>)</td>
<td style="text-align: center;">95,366,600</td>
<td style="text-align: center;">34,821.34M</td>
<td style="text-align: center;">19.10</td>
<td style="text-align: center;">4.52</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DarkNet Ref (<a
href="https://github.com/pjreddie/darknet">‘Darknet: Open source neural
networks in C’</a>)</td>
<td style="text-align: center;">7,319,416</td>
<td style="text-align: center;">367.59M</td>
<td style="text-align: center;">38.58</td>
<td style="text-align: center;">17.18</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">DarkNet Tiny (<a
href="https://github.com/pjreddie/darknet">‘Darknet: Open source neural
networks in C’</a>)</td>
<td style="text-align: center;">1,042,104</td>
<td style="text-align: center;">500.85M</td>
<td style="text-align: center;">40.74</td>
<td style="text-align: center;">17.84</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DarkNet 53 (<a
href="https://github.com/pjreddie/darknet">‘Darknet: Open source neural
networks in C’</a>)</td>
<td style="text-align: center;">41,609,928</td>
<td style="text-align: center;">7,133.86M</td>
<td style="text-align: center;">21.75</td>
<td style="text-align: center;">5.64</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">SqueezeResNet1.1 (<a
href="https://arxiv.org/abs/1602.07360">‘SqueezeNet: AlexNet-level
accuracy with 50x fewer parameters and &lt;0.5MB model size’</a>)</td>
<td style="text-align: center;">1,235,496</td>
<td style="text-align: center;">352.02M</td>
<td style="text-align: center;">40.09</td>
<td style="text-align: center;">18.21</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SqueezeNet1.1 (<a
href="https://arxiv.org/abs/1602.07360">‘SqueezeNet: AlexNet-level
accuracy with 50x fewer parameters and &lt;0.5MB model size’</a>)</td>
<td style="text-align: center;">1,235,496</td>
<td style="text-align: center;">352.02M</td>
<td style="text-align: center;">39.31</td>
<td style="text-align: center;">17.72</td>
<td style="text-align: center;">2016</td>
</tr>
<tr class="even">
<td style="text-align: center;">ResAttNet-92 (<a
href="https://arxiv.org/abs/1704.06904">‘Residual Attention Network for
Image Classification’</a>)</td>
<td style="text-align: center;">51.3M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">19.5</td>
<td style="text-align: center;">4.8</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="odd">
<td style="text-align: center;">CondenseNet (G=C=8) (<a
href="https://arxiv.org/abs/1711.09224">‘CondenseNet: An Efficient
DenseNet using Learned Group Convolutions’</a>)</td>
<td style="text-align: center;">4.8M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">26.2</td>
<td style="text-align: center;">8.3</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="even">
<td style="text-align: center;">DPN-68 (<a
href="https://arxiv.org/abs/1707.01629">‘Dual Path Networks’</a>)</td>
<td style="text-align: center;">12,611,602</td>
<td style="text-align: center;">2,351.84M</td>
<td style="text-align: center;">23.24</td>
<td style="text-align: center;">6.79</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ShuffleNet x1.0 (g=1) (<a
href="https://arxiv.org/abs/1707.01083">‘ShuffleNet: An Extremely
Efficient Convolutional Neural Network for Mobile Devices’</a>)</td>
<td style="text-align: center;">1,531,936</td>
<td style="text-align: center;">148.13M</td>
<td style="text-align: center;">34.93</td>
<td style="text-align: center;">13.89</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="even">
<td style="text-align: center;">DiracNetV2-18 (<a
href="https://arxiv.org/abs/1706.00388">‘DiracNets: Training Very Deep
Neural Networks Without Skip-Connections’</a>)</td>
<td style="text-align: center;">11,511,784</td>
<td style="text-align: center;">1,796.62M</td>
<td style="text-align: center;">31.47</td>
<td style="text-align: center;">11.70</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DiracNetV2-34 (<a
href="https://arxiv.org/abs/1706.00388">‘DiracNets: Training Very Deep
Neural Networks Without Skip-Connections’</a>)</td>
<td style="text-align: center;">21,616,232</td>
<td style="text-align: center;">3,646.93M</td>
<td style="text-align: center;">28.75</td>
<td style="text-align: center;">9.93</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="even">
<td style="text-align: center;">SENet-16 (<a
href="https://arxiv.org/abs/1709.01507">‘Squeeze-and-Excitation
Networks’</a>)</td>
<td style="text-align: center;">31,366,168</td>
<td style="text-align: center;">5,081.30M</td>
<td style="text-align: center;">25.65</td>
<td style="text-align: center;">8.20</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SENet-154 (<a
href="https://arxiv.org/abs/1709.01507">‘Squeeze-and-Excitation
Networks’</a>)</td>
<td style="text-align: center;">115,088,984</td>
<td style="text-align: center;">20,745.78M</td>
<td style="text-align: center;">18.62</td>
<td style="text-align: center;">4.61</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="even">
<td style="text-align: center;">MobileNet (<a
href="https://arxiv.org/abs/1704.04861">‘MobileNets: Efficient
Convolutional Neural Networks for Mobile Vision Applications’</a>)</td>
<td style="text-align: center;">4,231,976</td>
<td style="text-align: center;">579.80M</td>
<td style="text-align: center;">26.61</td>
<td style="text-align: center;">8.95</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="odd">
<td style="text-align: center;">NASNet-A 4@1056 (<a
href="https://arxiv.org/abs/1707.07012">‘Learning Transferable
Architectures for Scalable Image Recognition’</a>)</td>
<td style="text-align: center;">5,289,978</td>
<td style="text-align: center;">584.90M</td>
<td style="text-align: center;">25.68</td>
<td style="text-align: center;">8.16</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="even">
<td style="text-align: center;">NASNet-A 6@4032(<a
href="https://arxiv.org/abs/1707.07012">‘Learning Transferable
Architectures for Scalable Image Recognition’</a>)</td>
<td style="text-align: center;">88,753,150</td>
<td style="text-align: center;">23,976.44M</td>
<td style="text-align: center;">18.14</td>
<td style="text-align: center;">4.21</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DLA-34 (<a
href="https://arxiv.org/abs/1707.06484">‘Deep Layer
Aggregation’</a>)</td>
<td style="text-align: center;">15,742,104</td>
<td style="text-align: center;">3,071.37M</td>
<td style="text-align: center;">25.36</td>
<td style="text-align: center;">7.94</td>
<td style="text-align: center;">2017</td>
</tr>
<tr class="even">
<td style="text-align: center;">AirNet50-1x64d (r=2) (<a
href="https://ieeexplore.ieee.org/document/8510896">‘Attention Inspiring
Receptive-Fields Network for Learning Invariant
Representations’</a>)</td>
<td style="text-align: center;">27.43M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">22.48</td>
<td style="text-align: center;">6.21</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">BAM-ResNet-50 (<a
href="https://arxiv.org/abs/1807.06514">‘BAM: Bottleneck Attention
Module’</a>)</td>
<td style="text-align: center;">25.92M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">23.68</td>
<td style="text-align: center;">6.96</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">CBAM-ResNet-50 (<a
href="https://arxiv.org/abs/1807.06521">‘CBAM: Convolutional Block
Attention Module’</a>)</td>
<td style="text-align: center;">28.1M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">23.02</td>
<td style="text-align: center;">6.38</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">1.0-SqNxt-23v5 (<a
href="https://arxiv.org/abs/1803.10615">‘SqueezeNext: Hardware-Aware
Neural Network Design’</a>)</td>
<td style="text-align: center;">921,816</td>
<td style="text-align: center;">285.82M</td>
<td style="text-align: center;">40.77</td>
<td style="text-align: center;">17.85</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">1.5-SqNxt-23v5 (<a
href="https://arxiv.org/abs/1803.10615">‘SqueezeNext: Hardware-Aware
Neural Network Design’</a>)</td>
<td style="text-align: center;">1,953,616</td>
<td style="text-align: center;">550.97M</td>
<td style="text-align: center;">33.81</td>
<td style="text-align: center;">13.01</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">2.0-SqNxt-23v5 (<a
href="https://arxiv.org/abs/1803.10615">‘SqueezeNext: Hardware-Aware
Neural Network Design’</a>)</td>
<td style="text-align: center;">3,366,344</td>
<td style="text-align: center;">897.60M</td>
<td style="text-align: center;">29.63</td>
<td style="text-align: center;">10.66</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">ShuffleNetV2 (<a
href="https://arxiv.org/abs/1807.11164">‘ShuffleNet V2: Practical
Guidelines for Efficient CNN Architecture Design’</a>)</td>
<td style="text-align: center;">2,278,604</td>
<td style="text-align: center;">149.72M</td>
<td style="text-align: center;">31.44</td>
<td style="text-align: center;">11.63</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">456-MENet-24×1(g=3) (<a
href="https://arxiv.org/abs/1803.09127">‘Merging and Evolution:
Improving Convolutional Neural Networks for Mobile
Applications’</a>)</td>
<td style="text-align: center;">5.3M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">28.4</td>
<td style="text-align: center;">9.8</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">FD-MobileNet (<a
href="https://arxiv.org/abs/1802.03750">‘FD-MobileNet: Improved
MobileNet with A Fast Downsampling Strategy’</a>)</td>
<td style="text-align: center;">2,901,288</td>
<td style="text-align: center;">147.46M</td>
<td style="text-align: center;">34.23</td>
<td style="text-align: center;">13.38</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">MobileNetV2 (<a
href="https://arxiv.org/abs/1801.04381">‘MobileNetV2: Inverted Residuals
and Linear Bottlenecks’</a>)</td>
<td style="text-align: center;">3,504,960</td>
<td style="text-align: center;">329.36M</td>
<td style="text-align: center;">26.97</td>
<td style="text-align: center;">8.87</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">IGCV3 (<a
href="https://arxiv.org/abs/1806.00178">‘IGCV3: Interleaved Low-Rank
Group Convolutions for Efficient Deep Neural Networks’</a>)</td>
<td style="text-align: center;">3.5M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">28.22</td>
<td style="text-align: center;">9.54</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DARTS (<a
href="https://arxiv.org/abs/1806.09055">‘DARTS: Differentiable
Architecture Search’</a>)</td>
<td style="text-align: center;">4.9M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">26.9</td>
<td style="text-align: center;">9.0</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">PNASNet-5 (<a
href="https://arxiv.org/abs/1712.00559">‘Progressive Neural Architecture
Search’</a>)</td>
<td style="text-align: center;">5.1M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">25.8</td>
<td style="text-align: center;">8.1</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">AmoebaNet-C (<a
href="https://arxiv.org/abs/1802.01548">‘Regularized Evolution for Image
Classifier Architecture Search’</a>)</td>
<td style="text-align: center;">5.1M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">24.3</td>
<td style="text-align: center;">7.6</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">MnasNet (<a
href="https://arxiv.org/abs/1807.11626">‘MnasNet: Platform-Aware Neural
Architecture Search for Mobile’</a>)</td>
<td style="text-align: center;">4,308,816</td>
<td style="text-align: center;">317.67M</td>
<td style="text-align: center;">31.58</td>
<td style="text-align: center;">11.74</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">IBN-Net50-a (<a
href="https://arxiv.org/abs/1807.09441">‘Two at Once: Enhancing Learning
andGeneralization Capacities via IBN-Net’</a>)</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">22.54</td>
<td style="text-align: center;">6.32</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">MarginNet (<a
href="http://papers.nips.cc/paper/7364-large-margin-deep-networks-for-classification.pdf">‘Large
Margin Deep Networks for Classification’</a>)</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">22.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">A^2 Net (<a
href="http://papers.nips.cc/paper/7318-a2-nets-double-attention-networks.pdf">‘A^2-Nets:
Double Attention Networks’</a>)</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">23.0</td>
<td style="text-align: center;">6.5</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="even">
<td style="text-align: center;">FishNeXt-150 (<a
href="http://papers.nips.cc/paper/7356-fishnet-a-versatile-backbone-for-image-region-and-pixel-level-prediction.pdf">‘FishNet:
A Versatile Backbone for Image, Region, and Pixel Level
Prediction’</a>)</td>
<td style="text-align: center;">26.2M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">21.5</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2018</td>
</tr>
<tr class="odd">
<td style="text-align: center;">Shape-ResNet (<a
href="https://arxiv.org/pdf/1811.12231v2.pdf">‘IMAGENET-TRAINED CNNS ARE
BIASED TOWARDS TEXTURE; INCREASING SHAPE BIAS IMPROVES ACCURACY AND
ROBUSTNESS’</a>)</td>
<td style="text-align: center;">25.5M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">23.28</td>
<td style="text-align: center;">6.72</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">SimCNN(k=3 train) (<a
href="https://arxiv.org/pdf/1812.11446.pdf">‘Greedy Layerwise Learning
Can Scale to ImageNet’</a>)</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">28.4</td>
<td style="text-align: center;">10.2</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SKNet-50 (<a
href="https://arxiv.org/pdf/1903.06586.pdf">‘Selective Kernel
Networks’</a>)</td>
<td style="text-align: center;">27.5M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">20.79</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">SRM-ResNet-50 (<a
href="https://arxiv.org/pdf/1903.10829.pdf">‘SRM : A Style-based
Recalibration Module for Convolutional Neural Networks’</a>)</td>
<td style="text-align: center;">25.62M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">22.87</td>
<td style="text-align: center;">6.49</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">EfficientNet-B0 (<a
href="http://proceedings.mlr.press/v97/tan19a/tan19a.pdf">‘EfficientNet:
Rethinking Model Scaling for Convolutional Neural Networks’</a>)</td>
<td style="text-align: center;">5,288,548</td>
<td style="text-align: center;">414.31M</td>
<td style="text-align: center;">24.77</td>
<td style="text-align: center;">7.52</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">EfficientNet-B7b (<a
href="http://proceedings.mlr.press/v97/tan19a/tan19a.pdf">‘EfficientNet:
Rethinking Model Scaling for Convolutional Neural Networks’</a>)</td>
<td style="text-align: center;">66,347,960</td>
<td style="text-align: center;">39,010.98M</td>
<td style="text-align: center;">15.94</td>
<td style="text-align: center;">3.22</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ProxylessNAS (<a
href="https://arxiv.org/pdf/1812.00332.pdf">‘PROXYLESSNAS: DIRECT NEURAL
ARCHITECTURE SEARCH ON TARGET TASK AND HARDWARE’</a>)</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">24.9</td>
<td style="text-align: center;">7.5</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">MixNet-L (<a
href="https://arxiv.org/abs/1907.09595">‘MixNet: Mixed Depthwise
Convolutional Kernels’</a>)</td>
<td style="text-align: center;">7.3M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">21.1</td>
<td style="text-align: center;">5.8</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ECA-Net50 (<a
href="https://arxiv.org/pdf/1910.03151v1.pdf">‘ECA-Net: Efficient
Channel Attention for Deep Convolutional Neural Networks’</a>)</td>
<td style="text-align: center;">24.37M</td>
<td style="text-align: center;">3.86G</td>
<td style="text-align: center;">22.52</td>
<td style="text-align: center;">6.32</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">ECA-Net101 (<a
href="https://arxiv.org/pdf/1910.03151v1.pdf">‘ECA-Net: Efficient
Channel Attention for Deep Convolutional Neural Networks’</a>)</td>
<td style="text-align: center;">7.3M</td>
<td style="text-align: center;">7.35G</td>
<td style="text-align: center;">21.35</td>
<td style="text-align: center;">5.66</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ACNet-Densenet121 (<a
href="https://arxiv.org/abs/1908.03930">‘ACNet: Strengthening the Kernel
Skeletons for Powerful CNN via Asymmetric Convolution Blocks’</a>)</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">24.18</td>
<td style="text-align: center;">7.23</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">LIP-ResNet-50 (<a
href="https://arxiv.org/abs/1908.04156">‘LIP: Local Importance-based
Pooling’</a>)</td>
<td style="text-align: center;">23.9M</td>
<td style="text-align: center;">5.33G</td>
<td style="text-align: center;">21.81</td>
<td style="text-align: center;">6.04</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">LIP-ResNet-101 (<a
href="https://arxiv.org/abs/1908.04156">‘LIP: Local Importance-based
Pooling’</a>)</td>
<td style="text-align: center;">42.9M</td>
<td style="text-align: center;">9.06G</td>
<td style="text-align: center;">20.67</td>
<td style="text-align: center;">5.40</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">LIP-DenseNet-BC-121 (<a
href="https://arxiv.org/abs/1908.04156">‘LIP: Local Importance-based
Pooling’</a>)</td>
<td style="text-align: center;">8.7M</td>
<td style="text-align: center;">4.13G</td>
<td style="text-align: center;">23.36</td>
<td style="text-align: center;">6.84</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">MuffNet_1.0 (<a
href="http://openaccess.thecvf.com/content_ICCVW_2019/papers/CEFRL/Chen_MuffNet_Multi-Layer_Feature_Federation_for_Mobile_Deep_Learning_ICCVW_2019_paper.pdf">‘MuffNet:
Multi-Layer Feature Federation for Mobile Deep Learning’</a>)</td>
<td style="text-align: center;">2.3M</td>
<td style="text-align: center;">146M</td>
<td style="text-align: center;">30.1</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">MuffNet_1.5 (<a
href="http://openaccess.thecvf.com/content_ICCVW_2019/papers/CEFRL/Chen_MuffNet_Multi-Layer_Feature_Federation_for_Mobile_Deep_Learning_ICCVW_2019_paper.pdf">‘MuffNet:
Multi-Layer Feature Federation for Mobile Deep Learning’</a>)</td>
<td style="text-align: center;">3.4M</td>
<td style="text-align: center;">300M</td>
<td style="text-align: center;">26.9</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ResNet-34-Bin-5 (<a
href="https://arxiv.org/abs/1904.11486">‘Making Convolutional Networks
Shift-Invariant Again’</a>)</td>
<td style="text-align: center;">21.8M</td>
<td style="text-align: center;">3,672.68M</td>
<td style="text-align: center;">25.80</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">ResNet-50-Bin-5 (<a
href="https://arxiv.org/abs/1904.11486">‘Making Convolutional Networks
Shift-Invariant Again’</a>)</td>
<td style="text-align: center;">25.5M</td>
<td style="text-align: center;">3,877.95M</td>
<td style="text-align: center;">22.96</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">MobileNetV2-Bin-5 (<a
href="https://arxiv.org/abs/1904.11486">‘Making Convolutional Networks
Shift-Invariant Again’</a>)</td>
<td style="text-align: center;">3,504,960</td>
<td style="text-align: center;">329.36M</td>
<td style="text-align: center;">27.50</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">FixRes ResNeXt101 WSL (<a
href="https://arxiv.org/abs/1906.06423">‘Fixing the train-test
resolution discrepancy’</a>)</td>
<td style="text-align: center;">829M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">13.6</td>
<td style="text-align: center;">2.0</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="odd">
<td style="text-align: center;">Noisy Student*(L2) (<a
href="https://arxiv.org/abs/1911.04252">‘Self-training with Noisy
Student improves ImageNet classification’</a>)</td>
<td style="text-align: center;">480M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">12.6</td>
<td style="text-align: center;">1.8</td>
<td style="text-align: center;">2019</td>
</tr>
<tr class="even">
<td style="text-align: center;">TResNet-M (<a
href="https://arxiv.org/abs/2003.13630">‘TResNet: High Performance
GPU-Dedicated Architecture’</a>)</td>
<td style="text-align: center;">29.4M</td>
<td style="text-align: center;">5.5G</td>
<td style="text-align: center;">19.3</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DA-NAS-C (<a
href="https://arxiv.org/abs/2003.12563v1">‘DA-NAS: Data Adapted Pruning
for Efficient Neural Architecture Search’</a>)</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">467M</td>
<td style="text-align: center;">23.8</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="even">
<td style="text-align: center;">ResNeSt-50 (<a
href="https://arxiv.org/abs/2004.08955">‘ResNeSt: Split-Attention
Networks’</a>)</td>
<td style="text-align: center;">27.5M</td>
<td style="text-align: center;">5.39G</td>
<td style="text-align: center;">18.87</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ResNeSt-101 (<a
href="https://arxiv.org/abs/2004.08955">‘ResNeSt: Split-Attention
Networks’</a>)</td>
<td style="text-align: center;">48.3M</td>
<td style="text-align: center;">10.2G</td>
<td style="text-align: center;">17.73</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="even">
<td style="text-align: center;">ResNet-50-FReLU (<a
href="https://arxiv.org/abs/2007.11824v2">‘Funnel Activation for Visual
Recognition’</a>)</td>
<td style="text-align: center;">25.5M</td>
<td style="text-align: center;">3.87G</td>
<td style="text-align: center;">22.40</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ResNet-101-FReLU (<a
href="https://arxiv.org/abs/2007.11824v2">‘Funnel Activation for Visual
Recognition’</a>)</td>
<td style="text-align: center;">44.5M</td>
<td style="text-align: center;">7.6G</td>
<td style="text-align: center;">22.10</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="even">
<td style="text-align: center;">ResNet-50-MEALv2 (<a
href="https://arxiv.org/abs/2009.08453v1">‘MEAL V2: Boosting Vanilla
ResNet-50 to 80%+ Top-1 Accuracy on ImageNet without Tricks’</a>)</td>
<td style="text-align: center;">25.6M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">19.33</td>
<td style="text-align: center;">4.91</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ResNet-50-MEALv2 + CutMix (<a
href="https://arxiv.org/abs/2009.08453v1">‘MEAL V2: Boosting Vanilla
ResNet-50 to 80%+ Top-1 Accuracy on ImageNet without Tricks’</a>)</td>
<td style="text-align: center;">25.6M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">19.02</td>
<td style="text-align: center;">4.65</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="even">
<td style="text-align: center;">MobileNet V3-Large-MEALv2 (<a
href="https://arxiv.org/abs/2009.08453v1">‘MEAL V2: Boosting Vanilla
ResNet-50 to 80%+ Top-1 Accuracy on ImageNet without Tricks’</a>)</td>
<td style="text-align: center;">5.48M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">23.08</td>
<td style="text-align: center;">6.68</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="odd">
<td style="text-align: center;">EfficientNet-B0-MEALv2 (<a
href="https://arxiv.org/abs/2009.08453v1">‘MEAL V2: Boosting Vanilla
ResNet-50 to 80%+ Top-1 Accuracy on ImageNet without Tricks’</a>)</td>
<td style="text-align: center;">5.29M</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">21.71</td>
<td style="text-align: center;">6.05</td>
<td style="text-align: center;">2020</td>
</tr>
<tr class="even">
<td style="text-align: center;">T2T-ViT-7 (<a
href="https://arxiv.org/abs/2101.11986v1">‘Tokens-to-Token ViT: Training
Vision Transformers from Scratch on ImageNet’</a>)</td>
<td style="text-align: center;">4.2M</td>
<td style="text-align: center;">0.6G</td>
<td style="text-align: center;">28.8</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="odd">
<td style="text-align: center;">T2T-ViT-14 (<a
href="https://arxiv.org/abs/2101.11986v1">‘Tokens-to-Token ViT: Training
Vision Transformers from Scratch on ImageNet’</a>)</td>
<td style="text-align: center;">19.4M</td>
<td style="text-align: center;">4.8G</td>
<td style="text-align: center;">19.4</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="even">
<td style="text-align: center;">T2T-ViT-19 (<a
href="https://arxiv.org/abs/2101.11986v1">‘Tokens-to-Token ViT: Training
Vision Transformers from Scratch on ImageNet’</a>)</td>
<td style="text-align: center;">39.0M</td>
<td style="text-align: center;">8.0G</td>
<td style="text-align: center;">18.8</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="odd">
<td style="text-align: center;">NFNet-F0 (<a
href="https://arxiv.org/abs/2102.06171">‘High-Performance Large-Scale
Image Recognition Without Normalization’</a>)</td>
<td style="text-align: center;">71.5M</td>
<td style="text-align: center;">12.38G</td>
<td style="text-align: center;">16.4</td>
<td style="text-align: center;">3.2</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="even">
<td style="text-align: center;">NFNet-F1 (<a
href="https://arxiv.org/abs/2102.06171">‘High-Performance Large-Scale
Image Recognition Without Normalization’</a>)</td>
<td style="text-align: center;">132.6M</td>
<td style="text-align: center;">35.54G</td>
<td style="text-align: center;">15.4</td>
<td style="text-align: center;">2.9</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="odd">
<td style="text-align: center;">NFNet-F6+SAM (<a
href="https://arxiv.org/abs/2102.06171">‘High-Performance Large-Scale
Image Recognition Without Normalization’</a>)</td>
<td style="text-align: center;">438.4M</td>
<td style="text-align: center;">377.28G</td>
<td style="text-align: center;">13.5</td>
<td style="text-align: center;">2.1</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="even">
<td style="text-align: center;">EfficientNetV2-S (<a
href="https://arxiv.org/abs/2104.00298">‘EfficientNetV2: Smaller Models
and Faster Training’</a>)</td>
<td style="text-align: center;">24M</td>
<td style="text-align: center;">8.8G</td>
<td style="text-align: center;">16.1</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="odd">
<td style="text-align: center;">EfficientNetV2-M (<a
href="https://arxiv.org/abs/2104.00298">‘EfficientNetV2: Smaller Models
and Faster Training’</a>)</td>
<td style="text-align: center;">55M</td>
<td style="text-align: center;">24G</td>
<td style="text-align: center;">14.9</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="even">
<td style="text-align: center;">EfficientNetV2-L (<a
href="https://arxiv.org/abs/2104.00298">‘EfficientNetV2: Smaller Models
and Faster Training’</a>)</td>
<td style="text-align: center;">121M</td>
<td style="text-align: center;">53G</td>
<td style="text-align: center;">14.3</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="odd">
<td style="text-align: center;">EfficientNetV2-S (21k) (<a
href="https://arxiv.org/abs/2104.00298">‘EfficientNetV2: Smaller Models
and Faster Training’</a>)</td>
<td style="text-align: center;">24M</td>
<td style="text-align: center;">8.8G</td>
<td style="text-align: center;">15.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="even">
<td style="text-align: center;">EfficientNetV2-M (21k) (<a
href="https://arxiv.org/abs/2104.00298">‘EfficientNetV2: Smaller Models
and Faster Training’</a>)</td>
<td style="text-align: center;">55M</td>
<td style="text-align: center;">24G</td>
<td style="text-align: center;">13.9</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
<tr class="odd">
<td style="text-align: center;">EfficientNetV2-L (21k) (<a
href="https://arxiv.org/abs/2104.00298">‘EfficientNetV2: Smaller Models
and Faster Training’</a>)</td>
<td style="text-align: center;">121M</td>
<td style="text-align: center;">53G</td>
<td style="text-align: center;">13.2</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">2021</td>
</tr>
</tbody>
</table>
<h2 id="segmentation-models">Segmentation models</h2>
<table>
<colgroup>
<col style="width: 68%" />
<col style="width: 1%" />
<col style="width: 5%" />
<col style="width: 5%" />
<col style="width: 6%" />
<col style="width: 3%" />
<col style="width: 6%" />
</colgroup>
<thead>
<tr class="header">
<th style="text-align: center;">Model</th>
<th style="text-align: center;">Year</th>
<th style="text-align: center;">PASCAL-Context</th>
<th style="text-align: center;">Cityscapes (mIOU)</th>
<th style="text-align: center;">PASCAL VOC 2012 (mIOU)</th>
<th style="text-align: center;">COCO Stuff</th>
<th style="text-align: center;">ADE20K VAL (mIOU)</th>
</tr>
</thead>
<tbody>
<tr class="odd">
<td style="text-align: center;">U-Net (<a
href="https://arxiv.org/pdf/1505.04597.pdf">‘U-Net: Convolutional
Networks for Biomedical Image Segmentation’</a>)</td>
<td style="text-align: center;">2015</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">DeconvNet (<a
href="https://arxiv.org/pdf/1505.04366.pdf">‘Learning Deconvolution
Network for Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2015</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">72.5</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ParseNet (<a
href="https://arxiv.org/abs/1506.04579">‘ParseNet: Looking Wider to See
Better’</a>)</td>
<td style="text-align: center;">2015</td>
<td style="text-align: center;">40.4</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">69.8</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">Piecewise (<a
href="https://arxiv.org/abs/1504.01013">‘Efficient piecewise training of
deep structured models for semantic segmentation’</a>)</td>
<td style="text-align: center;">2015</td>
<td style="text-align: center;">43.3</td>
<td style="text-align: center;">71.6</td>
<td style="text-align: center;">78.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SegNet (<a
href="https://arxiv.org/pdf/1511.00561.pdf">‘SegNet: A Deep
Convolutional Encoder-Decoder Architecture for Image
Segmentation’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">56.1</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">FCN (<a
href="https://arxiv.org/pdf/1605.06211.pdf">‘Fully Convolutional
Networks for Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">37.8</td>
<td style="text-align: center;">65.3</td>
<td style="text-align: center;">62.2</td>
<td style="text-align: center;">22.7</td>
<td style="text-align: center;">29.39</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ENet (<a
href="https://arxiv.org/pdf/1606.02147.pdf">‘ENet: A Deep Neural Network
Architecture for Real-Time Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">58.3</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">DilatedNet (<a
href="https://arxiv.org/pdf/1511.07122.pdf">‘MULTI-SCALE CONTEXT
AGGREGATION BY DILATED CONVOLUTIONS’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">67.6</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">32.31</td>
</tr>
<tr class="odd">
<td style="text-align: center;">PixelNet (<a
href="https://arxiv.org/pdf/1609.06694.pdf">‘PixelNet: Towards a General
Pixel-Level Architecture’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">69.8</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">RefineNet (<a
href="https://arxiv.org/pdf/1611.06612.pdf">‘RefineNet: Multi-Path
Refinement Networks for High-Resolution Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">47.3</td>
<td style="text-align: center;">73.6</td>
<td style="text-align: center;">83.4</td>
<td style="text-align: center;">33.6</td>
<td style="text-align: center;">40.70</td>
</tr>
<tr class="odd">
<td style="text-align: center;">LRR (<a
href="https://arxiv.org/pdf/1605.02264.pdf">‘Laplacian Pyramid
Reconstruction and Refinement for Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">71.8</td>
<td style="text-align: center;">79.3</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">FRRN (<a
href="https://arxiv.org/pdf/1611.08323.pdf">‘Full-Resolution Residual
Networks for Semantic Segmentation in Street Scenes’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">71.8</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">MultiNet (<a
href="https://arxiv.org/pdf/1612.07695.pdf">‘MultiNet: Real-time Joint
Semantic Reasoning for Autonomous Driving’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">DeepLab (<a
href="https://arxiv.org/pdf/1606.00915.pdf">‘DeepLab: Semantic Image
Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully
Connected CRFs’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">45.7</td>
<td style="text-align: center;">64.8</td>
<td style="text-align: center;">79.7</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">LinkNet (<a
href="https://arxiv.org/pdf/1707.03718.pdf">‘LinkNet: Exploiting Encoder
Representations for Efficient Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">Tiramisu (<a
href="https://arxiv.org/pdf/1611.09326.pdf">‘The One Hundred Layers
Tiramisu: Fully Convolutional DenseNets for Semantic
Segmentation’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ICNet (<a
href="https://arxiv.org/pdf/1704.08545.pdf">‘ICNet for Real-Time
Semantic Segmentation on High-Resolution Images’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">70.6</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">ERFNet (<a
href="http://www.robesafe.uah.es/personal/eduardo.romera/pdfs/Romera17iv.pdf">‘Efficient
ConvNet for Real-time Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">68.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">PSPNet (<a
href="https://arxiv.org/pdf/1612.01105.pdf">‘Pyramid Scene Parsing
Network’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">47.8</td>
<td style="text-align: center;">80.2</td>
<td style="text-align: center;">85.4</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">44.94</td>
</tr>
<tr class="even">
<td style="text-align: center;">GCN (<a
href="https://arxiv.org/pdf/1703.02719.pdf">‘Large Kernel Matters —
Improve Semantic Segmentation by Global Convolutional Network’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">76.9</td>
<td style="text-align: center;">82.2</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">Segaware (<a
href="https://arxiv.org/pdf/1708.04607.pdf">‘Segmentation-Aware
Convolutional Networks Using Local Attention Masks’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">69.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">PixelDCN (<a
href="https://arxiv.org/pdf/1705.06820.pdf">‘PIXEL DECONVOLUTIONAL
NETWORKS’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">73.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DeepLabv3 (<a
href="https://arxiv.org/pdf/1706.05587.pdf">‘Rethinking Atrous
Convolution for Semantic Image Segmentation’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">85.7</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">DUC, HDC (<a
href="https://arxiv.org/pdf/1702.08502.pdf">‘Understanding Convolution
for Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">77.1</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ShuffleSeg (<a
href="https://arxiv.org/pdf/1803.03816.pdf">‘SHUFFLESEG: REAL-TIME
SEMANTIC SEGMENTATION NETWORK’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">59.3</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">AdaptSegNet (<a
href="https://arxiv.org/pdf/1802.10349.pdf">‘Learning to Adapt
Structured Output Space for Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">46.7</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">TuSimple-DUC (<a
href="https://arxiv.org/pdf/1702.08502.pdf">‘Understanding Convolution
for Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">80.1</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">83.1</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">R2U-Net (<a
href="https://arxiv.org/pdf/1802.06955.pdf">‘Recurrent Residual
Convolutional Neural Network based on U-Net (R2U-Net) for Medical Image
Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">Attention U-Net (<a
href="https://arxiv.org/pdf/1804.03999.pdf">‘Attention U-Net: Learning
Where to Look for the Pancreas’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">DANet (<a
href="https://arxiv.org/pdf/1809.02983.pdf">‘Dual Attention Network for
Scene Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">52.6</td>
<td style="text-align: center;">81.5</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">39.7</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ENCNet (<a
href="https://arxiv.org/abs/1803.08904">‘Context Encoding for Semantic
Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">51.7</td>
<td style="text-align: center;">75.8</td>
<td style="text-align: center;">85.9</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">44.65</td>
</tr>
<tr class="even">
<td style="text-align: center;">ShelfNet (<a
href="https://arxiv.org/pdf/1811.11254.pdf">‘ShelfNet for Real-time
Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">48.4</td>
<td style="text-align: center;">75.8</td>
<td style="text-align: center;">84.2</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">LadderNet (<a
href="https://arxiv.org/pdf/1810.07810.pdf">‘LADDERNET: MULTI-PATH
NETWORKS BASED ON U-NET FOR MEDICAL IMAGE SEGMENTATION’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">CCC-ERFnet (<a
href="https://arxiv.org/pdf/1812.04920v1.pdf">‘Concentrated-Comprehensive
Convolutions for lightweight semantic segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">69.01</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DifNet-101 (<a
href="http://papers.nips.cc/paper/7435-difnet-semantic-segmentation-by-diffusion-networks.pdf">‘DifNet:
Semantic Segmentation by Diffusion Networks’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">45.1</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">73.2</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">BiSeNet(Res18) (<a
href="https://arxiv.org/pdf/1808.00897.pdf">‘BiSeNet: Bilateral
Segmentation Network for Real-time Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">74.7</td>
<td style="text-align: center;">28.1</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">ESPNet (<a
href="https://arxiv.org/pdf/1803.06815.pdf">‘ESPNet: Efficient Spatial
Pyramid of Dilated Convolutions for Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">63.01</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">SPADE (<a
href="https://arxiv.org/pdf/1903.07291.pdf">‘Semantic Image Synthesis
with Spatially-Adaptive Normalization’</a>)</td>
<td style="text-align: center;">2019</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">62.3</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">37.4</td>
<td style="text-align: center;">38.5</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SeamlessSeg (<a
href="https://arxiv.org/pdf/1905.01220v1.pdf">‘Seamless Scene
Segmentation’</a>)</td>
<td style="text-align: center;">2019</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">77.5</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">EMANet (<a
href="https://arxiv.org/pdf/1907.13426.pdf">‘Expectation-Maximization
Attention Networks for Semantic Segmentation’</a>)</td>
<td style="text-align: center;">2019</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">88.2</td>
<td style="text-align: center;">39.9</td>
<td style="text-align: center;">?</td>
</tr>
</tbody>
</table>
<h2 id="detection-models">Detection models</h2>
<table>
<colgroup>
<col style="width: 79%" />
<col style="width: 2%" />
<col style="width: 6%" />
<col style="width: 6%" />
<col style="width: 3%" />
</colgroup>
<thead>
<tr class="header">
<th style="text-align: center;">Model</th>
<th style="text-align: center;">Year</th>
<th style="text-align: center;">VOC07 (mAP@IoU=0.5)</th>
<th style="text-align: center;">VOC12 (mAP@IoU=0.5)</th>
<th style="text-align: center;">COCO (mAP)</th>
</tr>
</thead>
<tbody>
<tr class="odd">
<td style="text-align: center;">R-CNN (<a
href="https://arxiv.org/pdf/1311.2524.pdf">‘Rich feature hierarchies for
accurate object detection and semantic segmentation’</a>)</td>
<td style="text-align: center;">2014</td>
<td style="text-align: center;">58.5</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">OverFeat (<a
href="https://arxiv.org/pdf/1312.6229.pdf">‘OverFeat: Integrated
Recognition, Localization and Detection using Convolutional
Networks’</a>)</td>
<td style="text-align: center;">2014</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">MultiBox (<a
href="https://www.cv-foundation.org/openaccess/content_cvpr_2014/papers/Erhan_Scalable_Object_Detection_2014_CVPR_paper.pdf">‘Scalable
Object Detection using Deep Neural Networks’</a>)</td>
<td style="text-align: center;">2014</td>
<td style="text-align: center;">29.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">SPP-Net (<a
href="https://arxiv.org/pdf/1406.4729.pdf">‘Spatial Pyramid Pooling in
Deep Convolutional Networks for Visual Recognition’</a>)</td>
<td style="text-align: center;">2014</td>
<td style="text-align: center;">59.2</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">MR-CNN (<a
href="https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Gidaris_Object_Detection_via_ICCV_2015_paper.pdf">‘Object
detection via a multi-region &amp; semantic segmentation-aware CNN
model’</a>)</td>
<td style="text-align: center;">2015</td>
<td style="text-align: center;">78.2</td>
<td style="text-align: center;">73.9</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">AttentionNet (<a
href="https://arxiv.org/pdf/1506.07704.pdf">‘AttentionNet: Aggregating
Weak Directions for Accurate Object Detection’</a>)</td>
<td style="text-align: center;">2015</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">Fast R-CNN (<a
href="https://arxiv.org/pdf/1504.08083.pdf">‘Fast R-CNN’</a>)</td>
<td style="text-align: center;">2015</td>
<td style="text-align: center;">70.0</td>
<td style="text-align: center;">68.4</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">Fast R-CNN (<a
href="https://papers.nips.cc/paper/5638-faster-r-cnn-towards-real-time-object-detection-with-region-proposal-networks.pdf">‘Faster
R-CNN: Towards Real-Time Object Detection with Region Proposal
Networks’</a>)</td>
<td style="text-align: center;">2015</td>
<td style="text-align: center;">73.2</td>
<td style="text-align: center;">70.4</td>
<td style="text-align: center;">36.8</td>
</tr>
<tr class="odd">
<td style="text-align: center;">YOLO v1 (<a
href="https://arxiv.org/pdf/1506.02640.pdf">‘You Only Look Once:
Unified, Real-Time Object Detection’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">66.4</td>
<td style="text-align: center;">57.9</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">G-CNN (<a
href="https://arxiv.org/pdf/1512.07729.pdf">‘G-CNN: an Iterative Grid
Based Object Detector’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">66.8</td>
<td style="text-align: center;">66.4</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">AZNet (<a
href="https://arxiv.org/pdf/1512.07711.pdf">‘Adaptive Object Detection
Using Adjacency and Zoom Prediction’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">70.4</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">22.3</td>
</tr>
<tr class="even">
<td style="text-align: center;">ION (<a
href="https://arxiv.org/pdf/1512.04143.pdf">‘Inside-Outside Net:
Detecting Objects in Context with Skip Pooling and Recurrent Neural
Networks’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">80.1</td>
<td style="text-align: center;">77.9</td>
<td style="text-align: center;">33.1</td>
</tr>
<tr class="odd">
<td style="text-align: center;">HyperNet (<a
href="https://arxiv.org/pdf/1604.00600.pdf">‘HyperNet: Towards Accurate
Region Proposal Generation and Joint Object Detection’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">76.3</td>
<td style="text-align: center;">71.4</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">OHEM (<a
href="https://arxiv.org/pdf/1604.03540.pdf">‘Training Region-based
Object Detectors with Online Hard Example Mining’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">78.9</td>
<td style="text-align: center;">76.3</td>
<td style="text-align: center;">22.4</td>
</tr>
<tr class="odd">
<td style="text-align: center;">MPN (<a
href="https://arxiv.org/pdf/1604.02135.pdf">‘A MultiPath Network for
Object Detection’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">33.2</td>
</tr>
<tr class="even">
<td style="text-align: center;">SSD (<a
href="https://arxiv.org/pdf/1512.02325.pdf">‘SSD: Single Shot MultiBox
Detector’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">76.8</td>
<td style="text-align: center;">74.9</td>
<td style="text-align: center;">31.2</td>
</tr>
<tr class="odd">
<td style="text-align: center;">GBDNet (<a
href="https://arxiv.org/pdf/1610.02579.pdf">‘Crafting GBD-Net for Object
Detection’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">77.2</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">27.0</td>
</tr>
<tr class="even">
<td style="text-align: center;">CPF (<a
href="https://pdfs.semanticscholar.org/40e7/4473cb82231559cbaeaa44989e9bbfe7ec3f.pdf">‘Contextual
Priming and Feedback for Faster R-CNN’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">76.4</td>
<td style="text-align: center;">72.6</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">MS-CNN (<a
href="https://arxiv.org/pdf/1607.07155.pdf">‘A Unified Multi-scale Deep
Convolutional Neural Network for Fast Object Detection’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">R-FCN (<a
href="https://arxiv.org/pdf/1605.06409.pdf">‘R-FCN: Object Detection via
Region-based Fully Convolutional Networks’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">79.5</td>
<td style="text-align: center;">77.6</td>
<td style="text-align: center;">29.9</td>
</tr>
<tr class="odd">
<td style="text-align: center;">PVANET (<a
href="https://arxiv.org/pdf/1608.08021.pdf">‘PVANET: Deep but
Lightweight Neural Networks for Real-time Object Detection’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">DeepID-Net (<a
href="https://arxiv.org/pdf/1412.5661.pdf">‘DeepID-Net: Deformable Deep
Convolutional Neural Networks for Object Detection’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">69.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">NoC (<a
href="https://arxiv.org/pdf/1504.06066.pdf">‘Object Detection Networks
on Convolutional Feature Maps’</a>)</td>
<td style="text-align: center;">2016</td>
<td style="text-align: center;">71.6</td>
<td style="text-align: center;">68.8</td>
<td style="text-align: center;">27.2</td>
</tr>
<tr class="even">
<td style="text-align: center;">DSSD (<a
href="https://arxiv.org/pdf/1701.06659.pdf">‘DSSD : Deconvolutional
Single Shot Detector’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">81.5</td>
<td style="text-align: center;">80.0</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">TDM (<a
href="https://arxiv.org/pdf/1612.06851.pdf">‘Beyond Skip Connections:
Top-Down Modulation for Object Detection’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">37.3</td>
</tr>
<tr class="even">
<td style="text-align: center;">FPN (<a
href="http://openaccess.thecvf.com/content_cvpr_2017/papers/Lin_Feature_Pyramid_Networks_CVPR_2017_paper.pdf">‘Feature
Pyramid Networks for Object Detection’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">36.2</td>
</tr>
<tr class="odd">
<td style="text-align: center;">YOLO v2 (<a
href="https://arxiv.org/pdf/1612.08242.pdf">‘YOLO9000: Better, Faster,
Stronger’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">78.6</td>
<td style="text-align: center;">73.4</td>
<td style="text-align: center;">21.6</td>
</tr>
<tr class="even">
<td style="text-align: center;">RON (<a
href="https://arxiv.org/pdf/1707.01691.pdf">‘RON: Reverse Connection
with Objectness Prior Networks for Object Detection’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">77.6</td>
<td style="text-align: center;">75.4</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DCN (<a
href="http://openaccess.thecvf.com/content_ICCV_2017/papers/Dai_Deformable_Convolutional_Networks_ICCV_2017_paper.pdf">‘Deformable
Convolutional Networks’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">DeNet (<a
href="https://arxiv.org/pdf/1703.10295.pdf">‘DeNet: Scalable Real-time
Object Detection with Directed Sparse Sampling’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">77.1</td>
<td style="text-align: center;">73.9</td>
<td style="text-align: center;">33.8</td>
</tr>
<tr class="odd">
<td style="text-align: center;">CoupleNet (<a
href="https://arxiv.org/pdf/1708.02863.pdf">‘CoupleNet: Coupling Global
Structure with Local Parts for Object Detection’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">82.7</td>
<td style="text-align: center;">80.4</td>
<td style="text-align: center;">34.4</td>
</tr>
<tr class="even">
<td style="text-align: center;">RetinaNet (<a
href="https://arxiv.org/pdf/1708.02002.pdf">‘Focal Loss for Dense Object
Detection’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">39.1</td>
</tr>
<tr class="odd">
<td style="text-align: center;">Mask R-CNN (<a
href="http://openaccess.thecvf.com/content_ICCV_2017/papers/He_Mask_R-CNN_ICCV_2017_paper.pdf">‘Mask
R-CNN’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">39.8</td>
</tr>
<tr class="even">
<td style="text-align: center;">DSOD (<a
href="https://arxiv.org/pdf/1708.01241.pdf">‘DSOD: Learning Deeply
Supervised Object Detectors from Scratch’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">77.7</td>
<td style="text-align: center;">76.3</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SMN (<a
href="http://openaccess.thecvf.com/content_ICCV_2017/papers/Chen_Spatial_Memory_for_ICCV_2017_paper.pdf">‘Spatial
Memory for Context Reasoning in Object Detection’</a>)</td>
<td style="text-align: center;">2017</td>
<td style="text-align: center;">70.0</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">YOLO v3 (<a
href="https://pjreddie.com/media/files/papers/YOLOv3.pdf">‘YOLOv3: An
Incremental Improvement’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">33.0</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SIN (<a
href="http://openaccess.thecvf.com/content_cvpr_2018/papers/Liu_Structure_Inference_Net_CVPR_2018_paper.pdf">‘Structure
Inference Net: Object Detection Using Scene-Level Context and
Instance-Level Relationships’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">76.0</td>
<td style="text-align: center;">73.1</td>
<td style="text-align: center;">23.2</td>
</tr>
<tr class="even">
<td style="text-align: center;">STDN (<a
href="http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhou_Scale-Transferrable_Object_Detection_CVPR_2018_paper.pdf">‘Scale-Transferrable
Object Detection’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">80.9</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">RefineDet (<a
href="http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhang_Single-Shot_Refinement_Neural_CVPR_2018_paper.pdf">‘Single-Shot
Refinement Neural Network for Object Detection’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">83.8</td>
<td style="text-align: center;">83.5</td>
<td style="text-align: center;">41.8</td>
</tr>
<tr class="even">
<td style="text-align: center;">MegDet (<a
href="http://openaccess.thecvf.com/content_cvpr_2018/papers/Peng_MegDet_A_Large_CVPR_2018_paper.pdf">‘MegDet:
A Large Mini-Batch Object Detector’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="odd">
<td style="text-align: center;">RFBNet (<a
href="https://arxiv.org/pdf/1711.07767.pdf">‘Receptive Field Block Net
for Accurate and Fast Object Detection’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">82.2</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
</tr>
<tr class="even">
<td style="text-align: center;">CornerNet (<a
href="https://arxiv.org/pdf/1808.01244.pdf">‘CornerNet: Detecting
Objects as Paired Keypoints’</a>)</td>
<td style="text-align: center;">2018</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">42.1</td>
</tr>
<tr class="odd">
<td style="text-align: center;">LibraRetinaNet (<a
href="https://arxiv.org/pdf/1904.02701v1.pdf">‘Libra R-CNN: Towards
Balanced Learning for Object Detection’</a>)</td>
<td style="text-align: center;">2019</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">43.0</td>
</tr>
<tr class="even">
<td style="text-align: center;">YOLACT-700 (<a
href="https://arxiv.org/pdf/1904.02689v1.pdf">‘YOLACT Real-time Instance
Segmentation’</a>)</td>
<td style="text-align: center;">2019</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">31.2</td>
</tr>
<tr class="odd">
<td style="text-align: center;">DetNASNet(3.8) (<a
href="https://arxiv.org/pdf/1903.10979v2.pdf">‘DetNAS: Backbone Search
for Object Detection’</a>)</td>
<td style="text-align: center;">2019</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">42.0</td>
</tr>
<tr class="even">
<td style="text-align: center;">YOLOv4 (<a
href="https://arxiv.org/pdf/2004.10934.pdf">‘YOLOv4: Optimal Speed and
Accuracy of Object Detection’</a>)</td>
<td style="text-align: center;">2020</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">46.7</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SOLO (<a
href="https://arxiv.org/pdf/1912.04488v3.pdf">‘SOLO: Segmenting Objects
by Locations’</a>)</td>
<td style="text-align: center;">2020</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">37.8</td>
</tr>
<tr class="even">
<td style="text-align: center;">D-SOLO (<a
href="https://arxiv.org/pdf/1912.04488v3.pdf">‘SOLO: Segmenting Objects
by Locations’</a>)</td>
<td style="text-align: center;">2020</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">40.5</td>
</tr>
<tr class="odd">
<td style="text-align: center;">SNIPER (<a
href="https://arxiv.org/pdf/2102.05646v1.pdf">‘Scale Normalized Image
Pyramids with AutoFocus for Object Detection’</a>)</td>
<td style="text-align: center;">2021</td>
<td style="text-align: center;">86.6</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">47.9</td>
</tr>
<tr class="even">
<td style="text-align: center;">AutoFocus (<a
href="https://arxiv.org/pdf/2102.05646v1.pdf">‘Scale Normalized Image
Pyramids with AutoFocus for Object Detection’</a>)</td>
<td style="text-align: center;">2021</td>
<td style="text-align: center;">85.8</td>
<td style="text-align: center;">?</td>
<td style="text-align: center;">47.9</td>
</tr>
</tbody>
</table>
Reference in New Issue View Git Blame Copy Permalink