ProxyCLIP: Proxy Attention Improves CLIP for Open-Vocabulary Segmentation

Benchmark Model Rank Results
unsupervised-semantic-segmentation-with-language-image-pre-training-on-cityscapes-valProxyCLIP#3mIoU: 42.0
unsupervised-semantic-segmentation-with-language-image-pre-training-on-coco-objectProxyCLIP#4mIoU: 39.2
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascal-context-59ProxyCLIP#4mIoU: 39.6
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascal-context-60ProxyCLIP#4mIoU: 35.4
unsupervised-semantic-segmentation-with-language-image-pre-training-on-ade20kProxyCLIP#5Mean IoU (val): 24.2
unsupervised-semantic-segmentation-with-language-image-pre-training-on-coco-stuff-171ProxyCLIP#6mIoU: 26.8
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascal-vocProxyCLIP#6mIoU: 65.0
unsupervised-semantic-segmentation-with-language-image-pre-training-on-pascalvoc-20ProxyCLIP#7mIoU: 83.3