In this work, we propose a deep learning-based approach for kin verification using a unified multi-task learning scheme where all kinship classes are jointly learned. This allows us to better utilize small training sets that are typical of kin verification. We introduce a novel approach for fusing the embeddings of kin images, to avoid overfitting, which is a common issue in training such networks. An adaptive sampling scheme is derived for the training set images, to resolve the inherent imbalance in kin verification datasets. A thorough ablation study exemplifies the effectivity of our approach, which is experimentally shown to outperform contemporary state-of-the-art kin verification results when applied to the Families In the Wild, FG2018, and FG2020 datasets.

We consider a problem which is an extension of the multi-terminal source coding problem with encoder breakdown studied by Berger-Yeung. In our problem when, and only when, encoder breakdown happens the decoder gets help from a third node in the form of side-information which is degraded with respect to the un-encoded source. Inner and outer bounds on the rate-distortion region are derived. The achievable region is tight when both source components are reconstructed losslessly at the non-breakdown scenario and partially tight when the distortion measures are degraded and deterministic. Our results also explain the missing sum-rate constraint in the characterization of the rate-distortion region of the Berger-Yeung problem.