Skip to content

Latest commit

 

History

History

deeplabv3

Rethinking Atrous Convolution for Semantic Image Segmentation

Reference

Chen, Liang-Chieh, George Papandreou, Florian Schroff, and Hartwig Adam. "Rethinking Atrous Convolution for Semantic Image Segmentation." arXiv preprint arXiv:1706.05587 (2017).

Performance

Cityscapes

Model Backbone Resolution Training Iters mIoU mIoU (flip) mIoU (ms+flip) Links
DeepLabV3 ResNet50_OS8 1024x512 80000 79.90% 80.22% 80.89% model | log | vdl
DeepLabV3 ResNet101_OS8 1024x512 80000 80.85% 81.09% 81.78% model | log | vdl

Pascal VOC 2012 + Aug

Model Backbone Resolution Training Iters mIoU mIoU (flip) mIoU (ms+flip) Links
DeepLabV3 ResNet50_OS8 512x512 40000 79.76% 80.11% 81.05% model | log | vdl
DeepLabV3 ResNet101_OS8 512x512 40000 80.62% 80.87% 81.59% model | log | vdl