Self-Supervised Spatial Correspondence Across Modalities
This addresses the challenge of aligning multimodal data without labeled pairs, which is useful for applications in computer vision and robotics.
The paper tackles the problem of finding cross-modal spatial correspondences between images from different visual modalities, such as RGB and depth, by extending a contrastive random walk framework to learn cycle-consistent features without explicit photo-consistency assumptions, achieving strong performance across geometric and semantic benchmarks.
We present a method for finding cross-modal space-time correspondences. Given two images from different visual modalities, such as an RGB image and a depth map, our model identifies which pairs of pixels correspond to the same physical points in the scene. To solve this problem, we extend the contrastive random walk framework to simultaneously learn cycle-consistent feature representations for both cross-modal and intra-modal matching. The resulting model is simple and has no explicit photo-consistency assumptions. It can be trained entirely using unlabeled data, without the need for any spatially aligned multimodal image pairs. We evaluate our method on both geometric and semantic correspondence tasks. For geometric matching, we consider challenging tasks such as RGB-to-depth and RGB-to-thermal matching (and vice versa); for semantic matching, we evaluate on photo-sketch and cross-style image alignment. Our method achieves strong performance across all benchmarks.