Location-aware Upsampling for Semantic Segmentation

13 Nov 2019  ·  Xiangyu He, Zitao Mo, Qiang Chen, Anda Cheng, Peisong Wang, Jian Cheng ·

Many successful learning targets such as minimizing dice loss and cross-entropy loss have enabled unprecedented breakthroughs in segmentation tasks. Beyond these semantic metrics, this paper aims to introduce location supervision into semantic segmentation. Based on this idea, we present a Location-aware Upsampling (LaU) that adaptively refines the interpolating coordinates with trainable offsets. Then, location-aware losses are established by encouraging pixels to move towards well-classified locations. An LaU is offset prediction coupled with interpolation, which is trained end-to-end to generate confidence score at each position from coarse to fine. Guided by location-aware losses, the new module can replace its plain counterpart (\textit{e.g.}, bilinear upsampling) in a plug-and-play manner to further boost the leading encoder-decoder approaches. Extensive experiments validate the consistent improvement over the state-of-the-art methods on benchmark datasets. Our code is available at https://github.com/HolmesShuan/Location-aware-Upsampling-for-Semantic-Segmentation

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Semantic Segmentation ADE20K LaU-regression-loss Validation mIoU 45.02 # 190
Test Score 56.32 # 4
Semantic Segmentation ADE20K LaU-offset-loss Validation mIoU 44.55 # 195
Test Score 56.41 # 3
Semantic Segmentation ADE20K val LaU-regression-loss mIoU 45.02 # 80
Semantic Segmentation PASCAL Context LaU-regression-loss (ResNet-101) mIoU 53.9 # 35

Methods