DiSa: Saliency-Aware Foreground-Background Disentangled Framework for Open-Vocabulary Semantic Segmentation

Benchmark Model Rank Results
open-vocabulary-semantic-segmentation-on-ade20k-150DiSa (CLIP ViT-B/16)–mIoU: 33.7
open-vocabulary-semantic-segmentation-on-ade20k-150DiSa (CLIP ViT-L/14)–mIoU: 38.9
open-vocabulary-semantic-segmentation-on-ade20k-847DiSa (CLIP ViT-B/16)–mIoU: 12.6
open-vocabulary-semantic-segmentation-on-ade20k-847DiSa (CLIP ViT-L/14)–mIoU: 16.3
open-vocabulary-semantic-segmentation-on-pascal-context-459DiSa (CLIP ViT-B/16)–mIoU: 20.3
open-vocabulary-semantic-segmentation-on-pascal-context-459DiSa (CLIP ViT-L/14)–mIoU: 24.9
open-vocabulary-semantic-segmentation-on-pascal-context-59DiSa (CLIP ViT-B/16)–mIoU: 59.3
open-vocabulary-semantic-segmentation-on-pascal-context-59DiSa (CLIP ViT-L/14)–mIoU: 64.7
open-vocabulary-semantic-segmentation-on-pascalvoc-20DiSa–mIoU: 98.7