Open-Vocabulary Semantic Segmentation with Mask-adapted CLIP

Benchmark Model Rank Results
open-vocabulary-semantic-segmentation-on-1OVSeg Swin-B#14mIoU: 55.7
open-vocabulary-semantic-segmentation-on-2OVSeg Swin-B#15mIoU: 29.6
open-vocabulary-semantic-segmentation-on-3OVSeg Swin-B#16mIoU: 9
open-vocabulary-semantic-segmentation-on-5OVSeg Swin-B#8mIoU: 94.5
open-vocabulary-semantic-segmentation-on-7OVSeg Swin-B#13mIoU: 12.4
semantic-segmentation-on-replicaOVSeg#4mIoU: 20.7