The paper "Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation" by Liang-Chieh Chen et al. proposes a novel model, DeepLabv3+, which combines the strengths of spatial pyramid pooling and encoder-decoder structures for semantic image segmentation. DeepLabv3+ extends DeepLabv3 by adding a decoder module to refine segmentation results, particularly along object boundaries. The authors also explore the use of depthwise separable convolution in both the atrous spatial pyramid pooling (ASPP) and decoder modules, improving both speed and accuracy. The model is evaluated on the PASCAL VOC 2012 and Cityscapes datasets, achieving test set performances of 89.0% and 82.1%, respectively, without any post-processing. The paper includes a detailed analysis of design choices and model variants, and provides a publicly available Tensorflow implementation.The paper "Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation" by Liang-Chieh Chen et al. proposes a novel model, DeepLabv3+, which combines the strengths of spatial pyramid pooling and encoder-decoder structures for semantic image segmentation. DeepLabv3+ extends DeepLabv3 by adding a decoder module to refine segmentation results, particularly along object boundaries. The authors also explore the use of depthwise separable convolution in both the atrous spatial pyramid pooling (ASPP) and decoder modules, improving both speed and accuracy. The model is evaluated on the PASCAL VOC 2012 and Cityscapes datasets, achieving test set performances of 89.0% and 82.1%, respectively, without any post-processing. The paper includes a detailed analysis of design choices and model variants, and provides a publicly available Tensorflow implementation.