RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentation

Benchmark Model Rank Results
semantic-segmentation-on-ade20kRefineNet#209Validation mIoU: 40.7
semantic-segmentation-on-ade20k-valRefineNet (ResNet-152)#85mIoU: 40.70
semantic-segmentation-on-ade20k-valRefineNet (ResNet-101)#86mIoU: 40.20
semantic-segmentation-on-cityscapesRefineNet (ResNet-101)#61Mean IoU (class): 73.6%
semantic-segmentation-on-coco-stuff-testRefineNet (ResNet-101)#16mIoU: 33.6%
semantic-segmentation-on-nyu-depth-v2RefineNet (ResNet-101)#63Mean IoU: 46.5%
semantic-segmentation-on-pascal-contextRefineNet#46mIoU: 47.3
semantic-segmentation-on-pascal-voc-2012Multipath-RefineNet#13Mean IoU: 84.2%
semantic-segmentation-on-trans10kRefineNet#13mIoU: 58.18%GFLOPs: 44.56