CNN backbones

结构介绍

• LeNet5(1998)
• AlexNet8(2012)
• VGG(2014)
  • VGG16
  • VGG19
• GoogLeNet22(2014)
  • Inception v1-v4
• ResNet(2015)
  • ResNet50
  • ResNet101
• DRN-26

Semantic Segmentation

• $p_{ij}$: 被预测为 $j$ 类样本的 $i$ 类样本
• Pixel Accuarcy: $\text{PA}=\frac{\sum _{i=0}^k{p}_{ii}}{\sum _{i=0}^k\sum _{j=0}^k{p}_{ij}}$
• Mean Pixel Accuracy: $\text{MPA}=\frac{1}{k}\sum _{i=0}^k\frac{p_{ii}}{\sum _{j=0}^k{p}_{ij}}$
• Mean Intersection over Union: $\text{MIoU}=\frac{1}{k}\sum _{i=0}^k\frac{p_{ii}}{\sum _{j=0}^k{p}_{ij}+\sum _{j=0}^k{p}_{ji}-p_{ii}}$
• Frequency Weighted Intersection over Union: $\text{FWIoU}=\frac{1}{\sum _{i=0}^k\sum _{j=0}^k{p}_{ij}}\sum _{i=0}\frac{p_{ii}\sum _{j=0}^k{p}_{ij}}{\sum _{j=0}^k{p}_{ij}+\sum _{j=0}^k{p}_{ji}-p_{ii}}$

架构

• FCN-8s
• DeepLab V2