Representation Separation for Semantic Segmentation with Vision Transformers

Benchmark Model Rank Results
semantic-segmentation-on-ade20kRSSeg-ViT-L (BEiT pretrain)–Validation mIoU: 58.4Params (M): 330
semantic-segmentation-on-ade20k-valRSSeg-ViT-L(BEiT pretrain)–mIoU: 58.4
semantic-segmentation-on-coco-stuff-testRSSeg-ViT-L–mIoU: 52.0%
semantic-segmentation-on-coco-stuff-testRSSeg-ViT-L (BEiT pretrain)–mIoU: 52.6%
semantic-segmentation-on-pascal-contextRSSeg-ViT-L–mIoU: 67.5
semantic-segmentation-on-pascal-contextRSSeg-ViT-L (BEiT pretrain)–mIoU: 68.9