Jamais Vu: Exposing the Generalization Gap in Supervised Semantic Correspondence

The goal of semantic correspondence (SC) estimation is to establish semantically meaningful matches across different instances of an object category. In this work, we illustrate how recent supervised SC methods generalize poorly beyond the annotated keypoints seen during training, thus effectively acting as keypoint detectors.

To address this, we propose a new approach for learning dense correspondences by lifting 2D keypoints into a canonical 3D space using monocular depth estimation.

Our method constructs a continuous canonical manifold that captures object geometry without requiring explicit 3D supervision or camera annotations. Additionally, we introduce SPair-U, an extension of SPair-71k with novel keypoint annotations, to better assess generalization. Experiments not only demonstrate that our model significantly outperforms supervised baselines on unseen keypoints, highlighting its effectiveness in learning robust correspondences, but that unsupervised baselines outperform supervised counterparts when evaluated across different datasets.