Figure 1. Comparative analysis of the VFM and the VLM fea-
tures. VFM: Visualization of PCA-computed features from DI-
NOv2 (the first three components of PCA, computed on the image
features, serve as color channels), displaying fine-grained details
but lacking text alignment. VLM: Image-text similarity map from
EVA02-CLIPusingthequery‘car’,demonstratinggoodalignment
with text but insufficient localization of queried objects. MFuser:
OurproposedfusionframeworkintegratesVFMandVLM,result-
ing in unified features that exhibit both precise locality and robust
text alignment. Quantitative results on synthetic-to-real DGSS
benchmarks further validate our approach, with MFuser consis-
tently achieving the highest mIoU scores across all tasks.