A Simple Baseline for Open-Vocabulary Semantic Segmentation with Pre-trained Vision-language Model

Open paper
Benchmark Model Rank Results
open-vocabulary-semantic-segmentation-on-cityscapesSimSeg#2mIoU: 34.5
open-vocabulary-semantic-segmentation-on-coco-stuff-171ZSSeg#2HIoU: 37.8
zero-shot-semantic-segmentation-on-coco-stuffzsseg#6Transductive Setting hIoU: 41.5Inductive Setting hIoU: 36.3
zero-shot-semantic-segmentation-on-pascal-voczsseg#6Transductive Setting hIoU: 79.3Inductive Setting hIoU: 77.5
open-vocabulary-semantic-segmentation-on-pascal-context-59SimSeg#19mIoU: 47.7
open-vocabulary-semantic-segmentation-on-pascalvoc-20ZSSeg#19hIoU: 77.5
open-vocabulary-semantic-segmentation-on-ade20k-150SimSeg#21mIoU: 20.5
open-vocabulary-semantic-segmentation-on-ade20k-847SimSeg#21mIoU: 7

Rank counts only results with a code link.