TextRegion: Text-Aligned Region Tokens from Frozen Image-Text Models

Benchmark Model Rank Results
unsupervised-semantic-segmentation-with-language-image-pre-training-on-ade20kTextRegion#2Mean IoU (val): 27.3
unsupervised-semantic-segmentation-with-language-image-pre-training-on-coco-stuff-171TextRegion#2mIoU: 31.2
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascal-context-59TextRegion#2mIoU: 46.1
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascal-context-60TextRegion#2mIoU: 41.2
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascal-vocTextRegion#2mIoU: 73.1
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascalvoc-20TextRegion#2mIoU: 89.5