Do It Yourself: Learning Semantic Correspondence from Pseudo-Labels
This work addresses semantic matching challenges in computer vision, offering an incremental improvement by reducing annotation needs while enhancing performance.
The paper tackled the problem of semantic correspondence estimation in computer vision, which suffers from ambiguities for symmetric objects or repeated parts, by proposing a 3D-aware pseudo-labeling method that refines off-the-shelf features; it achieved a new state-of-the-art on SPair-71k with absolute gains of over 4% and over 7% compared to methods with similar supervision requirements.
Finding correspondences between semantically similar points across images and object instances is one of the everlasting challenges in computer vision. While large pre-trained vision models have recently been demonstrated as effective priors for semantic matching, they still suffer from ambiguities for symmetric objects or repeated object parts. We propose improving semantic correspondence estimation through 3D-aware pseudo-labeling. Specifically, we train an adapter to refine off-the-shelf features using pseudo-labels obtained via 3D-aware chaining, filtering wrong labels through relaxed cyclic consistency, and 3D spherical prototype mapping constraints. While reducing the need for dataset-specific annotations compared to prior work, we establish a new state-of-the-art on SPair-71k, achieving an absolute gain of over 4% and of over 7% compared to methods with similar supervision requirements. The generality of our proposed approach simplifies the extension of training to other data sources, which we demonstrate in our experiments.