Probabilistic Warp Consistency for Weakly-Supervised Semantic Correspondences

TL;DR

This paper introduces Probabilistic Warp Consistency, a weakly-supervised learning method for semantic matching that improves performance by directly supervising dense matching scores with a probabilistic framework, handling occlusion and background clutter.

Contribution

It presents a novel probabilistic learning objective for semantic matching that incorporates occlusion handling and improves performance across multiple architectures and benchmarks.

Findings

Sets new state-of-the-art on four semantic matching benchmarks.

Improves strongly-supervised methods when combined with keypoint annotations.

Effectively handles occlusion and background clutter in semantic matching.

Abstract

We propose Probabilistic Warp Consistency, a weakly-supervised learning objective for semantic matching. Our approach directly supervises the dense matching scores predicted by the network, encoded as a conditional probability distribution. We first construct an image triplet by applying a known warp to one of the images in a pair depicting different instances of the same object class. Our probabilistic learning objectives are then derived using the constraints arising from the resulting image triplet. We further account for occlusion and background clutter present in real image pairs by extending our probabilistic output space with a learnable unmatched state. To supervise it, we design an objective between image pairs depicting different object classes. We validate our method by applying it to four recent semantic matching architectures. Our weakly-supervised approach sets a new state-of-the-art on four challenging semantic matching benchmarks. Lastly, we demonstrate that our objective also brings substantial improvements in the strongly-supervised regime, when combined with keypoint annotations.