DeepLab statistics

Tool stats & trends

Looking to identify usage trends or leading experts?

DeepLab specifications


Unique identifier OMICS_27969
Name DeepLab
Software type Application/Script
Interface Command line interface
Restrictions to use None
Operating system Unix/Linux
Programming languages C++, MATLAB, Python
Parallelization CUDA
Computer skills Advanced
Version 2
Stability Stable
Maintained Yes




No version available


  • person_outline Liang-Chieh Chen
  • person_outline George Papandreou
  • person_outline Iasonas Kokkinos
  • person_outline Kevin Murphy
  • person_outline Alan Yuille

Additional information

Publication for DeepLab

DeepLab citations


Structural inference embedded adversarial networks for scene parsing

PLoS One
PMCID: 5896926
PMID: 29649294
DOI: 10.1371/journal.pone.0195114

[…] hout using the adversarial loss function, we train the generator only with the multi-class cross-entropy loss function in an end-to-end fashion. In the feature learning layer, we utilize the modified Deeplab networks [, ] to extract the hierarchical visual features from RGB images and the hierarchical geometric features from depth images for the reason that the Deeplab networks is able to extract […]


Deep Deconvolutional Neural Network for Target Segmentation of Nasopharyngeal Cancer in Planning Computed Tomography Images

PMCID: 5770734
PMID: 29376025
DOI: 10.3389/fonc.2017.00315

[…] the proposed DDNN algorithm can learn the semantic information from nasopharyngeal CT data and produce high-quality segmentation of the target. We compared the proposed architecture with the popular Deeplab v2 VGG-16 model. This comparison revealed that our method achieved better segmentation performance. Our DDNN method deployed a deeper encoder and decoder neural network, which used convolution […]


Fully automatic acute ischemic lesion segmentation in DWI using convolutional neural networks

PMCID: 5480013
PMID: 28664034
DOI: 10.1016/j.nicl.2017.06.016

[…] Although the DeconvNet () was selected as the basis CNN in the proposed EDD Net, other generic CNN architectures, including the U-Net (), the DeepLab () and the FCN (), aiming at image segmentation were used as baseline comparison. In this set of experiments, comparisons were among single networks rather than ensembles. The training inputs […]


A top down manner based DCNN architecture for semantic image segmentation

PLoS One
PMCID: 5365135
PMID: 28339486
DOI: 10.1371/journal.pone.0174508

[…] ation. We employ coarse semantic labels from current DCNN-based methods to help produce better superpixels, and conversely utilize better superpixels to improve semantic labels. The brilliant FCN and DeepLab-CRF are used as baselines to validate our architecture. Moreover, we test and prove that the two processes are both valuable and complementary, which demonstrates that more compact combination […]


Looking to check out a full list of citations?

DeepLab institution(s)
Google Inc., Menlo Park, CA, USA; University College London, London, UK; Departments of Cognitive Science and Computer Science, Johns Hopkins University, Baltimore, MD, USA
DeepLab funding source(s)
Supported in part by the ARO 62250-CS, FP7-RECONFIG, FP7-MOBOT, and H2020-ISUPPORT EU projects.

DeepLab reviews

star_border star_border star_border star_border star_border
star star star star star

Be the first to review DeepLab