TagAlign: Improving Vision-Language Alignment with Multi-Tag Classification

Benchmark Model Rank Results
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascal-context-59TagAlign#5mIoU: 37.6
unsupervised-semantic-segmentation-with-language-image-pre-training-on-cityscapes-valTagAlign#6mIoU: 27.5
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascalvoc-20TagAlign#6mIoU: 87.9
unsupervised-semantic-segmentation-with-language-image-pre-training-on-ade20kTagAlign#7Mean IoU (val): 17.3
unsupervised-semantic-segmentation-with-language-image-pre-training-on-coco-objectTagAlign#7mIoU: 33.3
unsupervised-semantic-segmentation-with-language-image-pre-training-on-coco-stuff-171TagAlign#7mIoU: 25.3
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascal-vocTagAlign#9mIoU: 53.9
open-vocabulary-semantic-segmentation-on-pascalvoc-20TagAlign(trained with image-text pairs)#14mIoU: 87.9
open-vocabulary-semantic-segmentation-on-pascal-context-59TaAlign(trained with image-text pairs)#21mIoU: 37.6