Deeper, Broader and Artier Domain Generalization

TLDR

Domain generalization seeks to learn domain‑agnostic models from multiple training domains so they can be applied to unseen domains, motivated by target domains with distinct characteristics and sparse data such as sketch images, yet current methods are mainly evaluated on photo‑only benchmarks where domain distinctiveness and data sparsity are minimal. The authors argue that these benchmarks are overly straightforward and demonstrate that simple deep‑learning baselines perform surprisingly well on them. They propose a low‑rank parameterized CNN for end‑to‑end DG learning and introduce a new benchmark dataset spanning photo, sketch, cartoon, and painting domains. Experiments show the new dataset is more relevant and presents a larger domain shift, and the proposed method outperforms existing DG alternatives, providing a more significant challenge for future research.

Abstract

The problem of domain generalization is to learn from multiple training domains, and extract a domain-agnostic model that can then be applied to an unseen domain. Domain generalization (DG) has a clear motivation in contexts where there are target domains with distinct characteristics, yet sparse data for training. For example recognition in sketch images, which are distinctly more abstract and rarer than photos. Nevertheless, DG methods have primarily been evaluated on photo-only benchmarks focusing on alleviating the dataset bias where both problems of domain distinctiveness and data sparsity can be minimal. We argue that these benchmarks are overly straightforward, and show that simple deep learning baselines perform surprisingly well on them. In this paper, we make two main contributions: Firstly, we build upon the favorable domain shift-robust properties of deep learning methods, and develop a low-rank parameterized CNN model for end-to-end DG learning. Secondly, we develop a DG benchmark dataset covering photo, sketch, cartoon and painting domains. This is both more practically relevant, and harder (bigger domain shift) than existing benchmarks. The results show that our method outperforms existing DG alternatives, and our dataset provides a more significant DG challenge to drive future research.