目标检测汇总(随心所欲更新)

Posted suckchen

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了目标检测汇总(随心所欲更新)相关的知识,希望对你有一定的参考价值。

梳理一遍,还需补充哪些欢迎指出。 指标 MS COCO

 

一、单阶段目标检测(基于anchor的)

   multi-scale backbone arxiv/ github AP AP50 AP75 APs    APM  APL 亮点 time
YOLO   DarkNet-53                  

SSD300(Google)

  VGG https://arxiv.org/pdf/1512.02325.pdf 23.2 41.2 23.4 5.3 23.2 38.6    
SSD512x512   VGG   26.8 46.5 27.8 9.0 28.9 41.9    

RetinaNet

FAIR facebook)

  ResNet101FPN https://arxiv.org/pdf/1708.02002.pdf 39.1 59.1 42.3 21.8 42.7 50.2   Focal loss 类别平衡损失  
    ResNeXt101   40.8 61.1 44.1 24.1 44.2 51.2    

RefineDet512

( 中科院)

  ResNet101 https://arxiv.org/pdf/1711.06897.pdf 36.4  57.5 39.5 16.6 39.9 51.4    
      41.8 62.9 45.7 25.6 45.1 54.1    

M2Det

(阿里达摩+北大)

  VGG16 https://arxiv.org/pdf/1811.04533.pdf 41.0  59.7 45.0 22.1 46.5 53.8    
512x512  VGG16   44.2 64.6 49.3 29.2  47.9 55.1     
800x800   ResNet101    38.8 59.4 41.7 20.5  43.9 53.4    
  ResNet101    43.9  64.4 48.4 29.6  49.6  54.3    
YOLO-v4 512X512   CSPDarknet-53 https://arxiv.org/pdf/2004.10934.pdf 43.0 64.9 46.5 24.3  46.1  55.2     
608   CSPDarknet-53   43.5 65.7 47.3 26.7 46.7 53.3    

 二、Anchor-Free

 

multi-scale arxiv backbone AP AP50 AP75 APS APM APL 亮点
CornetNet   https://arxiv.org/pdf/1808.01244.pdf Hourglass 104 40.5 56.5 43.1 19.4 42.7 53.9  
    Hourglass 104 42.1  57.8  45.3 20.8 44.8 56.7  
ExtremNet   https://arxiv.org/pdf/1901.08043.pdf Hourglass 104 40.2 55.5 43.2 20.4 43.2 53.1  
    Hourglass 104 43.7  60.5 47.0 24.1 46.9 57.6  
FSAF800   https://arxiv.org/pdf/1903.00621.pdf ResNext-101 42.9 63.8  46.3 26.6 46.2 52.7  
FSAF  √     44.6 65.2 48.6 29.7 47.1 54.6  
FCOS   https://arxiv.org/pdf/1904.01355.pdf Hourglass109 40.5 56.5 43.1 19.4 42.7 53.9  
      ResNeXt-64x4d-101-FPN 44.7 64.1  48.1 27.6 47.5 55.6  
FoveaBox   https://arxiv.org/pdf/1904.03797v1.pdf resnet101 40.6 60.1 43.5  23.3 45.2 54.5  
      ResNeXt101 42.1   61.9 45.2 24.9 46.8 55.6  
RPDet   https://arxiv.org/pdf/1904.11490.pdf ResNet101 41.0   62.9 44.3  23.6 44.1 51.7  
    ResNet101-DCN 43.5 61.3  46.7 25.3 45.3 55.0  
CenterNet:keypoint triplet   https://arxiv.org/pdf/1904.08189.pdf Hourglass 52 41.6  59.4 44.2 22.5 43.1 54.1  
      Hourglass 104 44.9  62.9 48.1 25.6 47.4 57.4  
    Hourglass 52   43.5  61.3 46.7 25.3 45.3 55.0  
    Hourglass 104 47.0  64.5 50.7 28.9 49.9 58.9  

 

 

三、二阶段目标检测:

  Faster RCNN

  DetNet  2018

  Cascade RCNN

   LibraRCNN

  efficientDet

  maskrcnn

 

 

 

四、轻量化的目标检测

  input size backbone arxiv   AP AP50 AP75  亮点
ThunderNet 320x320 SNet146 https://arxiv.org/pdf/1903.11752.pdf   23.6 40.2  24.5  
  320x320 SENet535     28.0 46.2 29.5  
                 

以上是关于目标检测汇总(随心所欲更新)的主要内容,如果未能解决你的问题,请参考以下文章

3D目标检测论文方法汇总 2022部分持续更新中~

目标检测相关论文和代码资源汇总

[人工智能-深度学习-67]:目标检测 - 常见目标检测算法大汇总

深度学习检测小目标常用方法汇总

航空遥感图像(Aerial Images)目标检测数据集汇总

(转载)航空遥感图像(Aerial Images)目标检测数据集汇总