Search papers, labs, and topics across Lattice.
This study investigates the ability of frozen DINOv3 features to establish a region-level facial coordinate system that distinguishes facial components across identities and time without specialized training. By evaluating cross-identity nearest-neighbor matching and temporal label propagation on 200 real videos, the authors demonstrate that DINOv3 achieves 83.0% semantic accuracy, significantly outperforming a random baseline and showing high temporal tracking accuracy. The findings highlight the effectiveness of intermediate layers in DINOv3 for dense facial analysis, suggesting that frozen models can provide robust representations for facial correspondence tasks.
DINOv3 achieves 83% accuracy in region-level facial correspondence, outperforming random baselines and revealing the power of frozen models in zero-shot settings.
Frozen self-supervised vision models can align parts of generic objects, but it remains unclear whether this correspondence extends to human faces, where global layout is shared while identity-specific appearance varies sharply. We test whether frozen DINOv3 features define a region-level facial coordinate system: a feature space in which eyes, brows, nose, mouth, skin, and hair remain distinguishable across people and across time without face-specific training. Using DINOv3 ViT-L/16 patch embeddings and FaRL only as a face-part labeling interface, we evaluate cross-identity nearest-neighbor matching and temporal label propagation on 200 CelebDF-v2 real videos. DINOv3 achieves 83.0% region-level semantic accuracy under unconstrained cross-identity matching, compared with a 23.0% area-weighted random baseline, and 95.5% temporal tracking accuracy without a learned temporal module. A no-FaRL control collapses to 0.9%, showing that FaRL supplies semantic initialization while DINOv3 supplies dense spatial correspondence. The strongest correspondence appears at an intermediate layer: block 18 gives a 4.93x same-region versus cross-region discrimination ratio, compared with 1.48x at the final block. Against CLIP ViT-L/14, DINOv3 shows only a small aggregate advantage but a +16.8 pp gain on anatomical regions, indicating that image-level contrastive supervision captures coarse facial layout but not fine-grained anatomical identity. These results establish frozen DINOv3 as a strong zero-shot representation for region-level facial correspondence and identify intermediate self-supervised features as the most useful layer for dense face analysis.
