Poster
Unsupervised Hierarchy-Agnostic Segmentation: Parsing Semantic Image Structure
Simone Rossetti · Fiora Pirri
East Exhibit Hall A-C #1502
Unsupervised semantic segmentation aims to discover groupings within images, capturing objects' view-invariance without external supervision. This task is inherently ambiguous due to the variable levels of granularity in natural groups. Existing methods often bypass this ambiguity using dataset-specific priors. In our research, we address this ambiguity head-on and provide a universal tool for pixel-level semantic parsing of images guided by the latent representations encoded in self-supervised models. We introduce a novel algebraic methodology that recursively identifies latent semantic regions, dynamically estimates the number of components, and ensures smoothness in the partitioning process. The innovative approach identifies scene-conditioned primitives within a dataset and creates a hierarchy-agnostic semantic regions tree of the image pixels. The model captures fine and coarse semantic details, producing a nuanced and unbiased segmentation. We present a new metric for estimating the quality of the semantic segmentation of discovered elements on different levels of the hierarchy. The metric validates the intrinsic nature of the compositional relations among parts, objects, and scenes in a hierarchy-agnostic domain. Our results prove the power of this methodology, uncovering semantic regions without prior definitions and scaling effectively across various datasets. This robust framework for unsupervised image segmentation proves more accurate semantic hierarchical relationships between scene elements than traditional algorithms. The experiments underscore its potential for broad applicability in image analysis tasks, showcasing its ability to deliver a detailed and unbiased segmentation that surpasses existing unsupervised methods.
Live content is unavailable. Log in and register to view live content