Semi-Supervised Semantic Matching
This addresses the challenge of limited training data for semantic matching in computer vision, offering an incremental improvement over existing methods.
The paper tackles the problem of correspondence estimation between semantically related images by proposing a semi-supervised learning framework with cyclic consistency constraints on unlabeled pairs, achieving state-of-the-art results on a benchmark dataset.
Convolutional neural networks (CNNs) have been successfully applied to solve the problem of correspondence estimation between semantically related images. Due to non-availability of large training datasets, existing methods resort to self-supervised or unsupervised training paradigm. In this paper we propose a semi-supervised learning framework that imposes cyclic consistency constraint on unlabeled image pairs. Together with the supervised loss the proposed model achieves state-of-the-art on a benchmark semantic matching dataset.