Semantic Segmentation can be defined as a classification task on the pixel level; an image consists of various pixels, and these pixels grouped together define different elements in an image, therefore a method of classifying these pixels into the elements is called semantic image segmentation.
0) Overview:
Dataset:
Metrics:
Purpose:
1) Papers:
Modern semantic segmentation approaches are pioneered by the fully convolutional network (FCN) (Long et al., 2015). Many methods have been proposed based on this method to improve the segmentation results, such as increasing the resolution of feature maps with dilated/atrous convolutions (Chen et al., 2017, 2018), enriching context information (Yuan et al., 2020a; Fu et al., 2019; Zhao et al., 2017), using an encoder-decoder architecture (Chen et al., 2018;Milletari et al., 2016; Ronneberger et al., 2015), or some refinement schemes (Yuan et al., 2020b; Krähenbühl & Koltun, 2011). See (Minaee et al., 2021b) for a comprehensive review of these approaches.
Paper List:
"Image segmentation using deep learning: A survey", 2020.
[FCNs] "Fully convolutional networks for semantic segmentation", 2015.
Increase Feature Maps Resolution
"Rethinking atrous convolution for semantic image segmentation", 2017.
"Encoder-decoder with atrous separable convolution for semantic image segmentation", 2018.
Enrich Context Information
"Dual attention network for scene segmentation", 2019.
"Objectcontextual representations for semantic segmentation", 2019.
"Co-occurrent features in semantic segmentation", 2020.
"Pyramid scene parsing network", 2017.
Use Encoder-Decoder Architecture
"Encoder-decoder with atrous separable convolution for semantic image segmentation", 2018.
"Panoptic feature pyramid networks", 2019.
"V-net: Fully convolutional neural networks for volumetric medical image segmentation", 2016.
"Unet: Convolutional networks for biomedical image segmentation", 2015.
Refinement
"Efficient inference in fully connected crfs with gaussian edge potentials", 2011.
"Iterative instance segmentation", 2016.
"Segfix: Model-agnostic boundary refinement for segmentation", 2016.
References:
