Skip to content

Latest commit

 

History

History

deeplabv3p

Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation

Reference

Chen, Liang-Chieh, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. "Encoder-decoder with atrous separable convolution for semantic image segmentation." In Proceedings of the European conference on computer vision (ECCV), pp. 801-818. 2018.

Performance

Cityscapes

Model Backbone Resolution Training Iters mIoU mIoU (flip) mIoU (ms+flip) Links
DeepLabV3P ResNet50_OS8 1024x512 80000 80.36% 80.57% 80.81% model | log | vdl
DeepLabV3P* ResNet50_OS8 1024x512 80000 81.18% 81.42% 81.48% model | log | vdl
DeepLabV3P ResNet101_OS8 1024x512 80000 81.10% 81.38% 81.24% model | log | vdl
DeepLabV3P ResNet101_OS8 769x769 80000 81.53% 81.88% 82.12% model | log | vdl

DeepLabV3P* is DeepLabV3P with RMI Loss, which requires paddlepaddle=2.2.

Pascal VOC 2012 + Aug

Model Backbone Resolution Training Iters mIoU mIoU (flip) mIoU (ms+flip) Links
DeepLabV3P ResNet50_OS8 512x512 40000 80.66% 81.33% 81.49% model | log | vdl
DeepLabV3P ResNet101_OS8 512x512 40000 80.60% 80.77% 80.75% model | log | vdl