Multi-Scale High-Resolution Vision Transformer for Semantic Segmentation

Benchmark Model Rank Results
semantic-segmentation-on-cityscapes-valHRViT-b3 (SegFormer, SS)#26mIoU: 83.16%
semantic-segmentation-on-cityscapes-valHRViT-b2 (SegFormer, SS)#27mIoU: 82.81%
semantic-segmentation-on-cityscapes-valHRViT-b1 (SegFormer, SS)#37mIoU: 81.63%
semantic-segmentation-on-ade20kHRViT-b3 (SegFormer, SS)#128Validation mIoU: 50.2Params (M): 28.7GFLOPs (512 x 512): 67.9
semantic-segmentation-on-ade20kHRViT-b2 (SegFormer, SS)#158Validation mIoU: 48.76Params (M): 20.8GFLOPs (512 x 512): 28.0
semantic-segmentation-on-ade20kHRViT-b1 (SegFormer, SS)#199Validation mIoU: 45.88Params (M): 8.2GFLOPs (512 x 512): 14.6