Semantic Hierarchy-Aware Segmentation

Abstract

Humans are able to recognize structured relations in observation, allowing us to decompose complex scenes into simpler parts and abstract the visual world at multiple levels. However, such hierarchical reasoning ability of human perception remains largely unexplored in current literature of semantic segmentation. Existing works are often aware of flatten labels and distinguish all the semantic categories exclusively for each pixel. In this work, we instead address hierarchical semantic segmentation (HSS), with the aim of providing a structured, pixel-wise description of visual observation in terms of a class hierarchy. We devise <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Hssn</small> , a general HSS framework that tackles two critical issues in this task: <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i)</b> how to efficiently adapt existing hierarchy-agnostic segmentation networks to the HSS setting, and <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ii)</b> how to leverage the class hierarchy to regularize HSS network learning. To address <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i)</b> , <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Hssn</small> directly casts HSS as a pixel-wise multi-label classification task, only bringing minimal architecture change to current segmentation models. To solve <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ii)</b> , <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Hssn</small> first explores inherent properties of the hierarchy as a training objective, which enforces segmentation predictions to obey the hierarchy structure. Furthermore, with a set of hierarchy-induced margin constraints, <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Hssn</small> efficiently reshapes the learned pixel embedding space, so as to generate hierarchy-aware pixel representations and facilitate structured segmentation eventually. Building upon <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Hssn</small> , we further exploit the mutual exclusion relation between semantic labels and strengthen the margin based regularization strategy with more meaningful constrains, leading to <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Hssn</small> +, a more effective framework for HSS. We conduct extensive experiments on six semantic segmentation datasets (i.e., Mapillary Vistas 2.0, Cityscapes, LIP, PASCAL-Person-Part, PASCAL-Part-58, and PASCAL-Part-108), with different class hierarchies, network architectures, and backbones, and the results confirm the generalization and superiority of our algorithms.