[Semantic Segmentation] Some Classic network for Semantic Segmentation

Semantic Segmentation

• FCN
• U-Net
• PSPNet

FCN

Our key insight is to build “fully convolutional”networks that take input of arbitrary size and produce correspondingly-sized output with efficient inference and learning.

We then define a skip architecture that combines semantic information from a deep, coarse layer with appearance information from a shallow, fine layer to produce accurate and detailed segmentations.

Structure:

Structure of FCN

The FCN will choose a classify network as backbone (e.g. VGGNet, ResNet). It abandon FC and replace them with Conv.

Details:

UNet

Structure:

Structure of UNet

PSPNet

Structure:

Structure of PSPNet