The evolution of the DeepLab circle of relatives is feature for the evolution of FCN impressed fashions for symbol segmentation. DeepLab variants can also be present in each, naive-decoder and encoder-decoder fashions. Hence, the guide orientates in this circle of relatives through first taking a look at naive-decoders after which turning in opposition to encoder-decoder fashions.
The maximum essential insights of naive-decoder fashions are principally the status quo of so known as atrous convolutions and lengthy vary symbol context exploitation for prediction on pixel degree. Atrous convolutions are a variant of ordinary convolutions, which permit an expanding receptive box with out the lack of symbol solution. The well-known Atrous Spatial Pyramid Pooling module (ASPP module) in DeepLab-V2  and later combines each: atrous convolutions and lengthy vary symbol context exploitation. When reading the next literature, center of attention at the trends of the ones options — Atrous convolutions, the ASPP module and lengthy vary symbol context exploitation/parsing.
Today, probably the most well-known encoder-decoder is most certainly the U-Net . A CNN which used to be advanced for examining scientific pictures. Its transparent construction invited many researchers to experiment and undertake it and it’s well-known for its skip connections, which permit the sharing of options between the encoder and decoder paths. Encoder-decoder fashions center of attention on bettering the semantically wealthy function maps all over upsampling within the decoder with extra in the community exact function maps from the encoder.
With the literature to hand, it is possible for you to to replicate on trendy symbol segmentation papers and implementations with CNNs. Let’s meet once more in Part III, the place we will be able to speak about object detection.
 Hoeser, T; Kuenzer, C. Object Detection and Image Segmentation with Deep Learning on Earth Observation Data: A Review-Part I: Evolution and Recent Trends. Remote Sensing 2020, 12(10), 1667. DOI: 10.3390/rs12101667.
 Long, J.; Shelhamer, E.; Darrell, T. Fully Convolutional Networks for Semantic Segmentation. IEEE Trans. Pattern Anal. Mach. Intell. 2014, 39, 640–651.
 Chen, L.C.; Zhu, Y.; Papandreou, G.; Schroff, F.; Adam, H. Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation. In Computer Vision–ECCV 2018; Ferrari, V., Hebert, M., Sminchisescu, C.; Weiss, Y., Eds.; Springer International Publishing: Cham, Switzerland, 2018; pp. 833–851
 Chen, L.C.; Papandreou, G.; Kokkinos, I.; Murphy, Okay.; Yuille, A.L. DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs. IEEE Trans. Pattern Anal.
Mach. Intell. 2016, 40, 834–848.
 Ronneberger, O.; Fischer, P.; Brox, T. U-Net: Convolutional Networks for Biomedical Image Segmentation. In Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015; Navab, N., Hornegger, J.,
Wells, W.M., Frangi, A.F., Eds.; Springer International Publishing: Cham, Switzerland, 2015; pp. 234–241.