OV-Stitcher improves training-free open-vocabulary semantic segmentation by stitching sub-image features to enable global attention in the final encoder block, raising mIoU from 48.7 to 50.7 across eight benchmarks.
The open images dataset v4: Unified image classification, object detection, and visual relationship detection at scale
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
OV-Stitcher: A Global Context-Aware Framework for Training-Free Open-Vocabulary Semantic Segmentation
OV-Stitcher improves training-free open-vocabulary semantic segmentation by stitching sub-image features to enable global attention in the final encoder block, raising mIoU from 48.7 to 50.7 across eight benchmarks.