unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning

2025-06-02

AI Summary

This paper addresses unsupervised multi-object segmentation from a single natural image—i.e., accurately segmenting multiple complex, overlapping objects in real-world scenes without any annotations, training, or reconstruction objectives. We propose a two-stage framework: Stage I jointly models center-boundary-shape priors via center-point detection, boundary-aware feature disentanglement, and geometric consistency constraints; Stage II performs fully inference-based (zero-parameter, zero-training) graph clustering and spatial relation reasoning for multi-object discovery. To our knowledge, this is the first method achieving robust dense segmentation on six real-world benchmarks—including COCO—outperforming prior art significantly in crowded scenes and establishing new state-of-the-art performance. Key contributions include: (1) the first center-boundary-cooperative unsupervised object representation; (2) a fully training-free multi-object inference mechanism; and (3) the first validated unsupervised segmentation paradigm effective on complex natural images.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Unsupervised & Self-Supervised Learning
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

We study the challenging problem of unsupervised multi-object segmentation on single images. Existing methods, which rely on image reconstruction objectives to learn objectness or leverage pretrained image features to group similar pixels, often succeed only in segmenting simple synthetic objects or discovering a limited number of real-world objects. In this paper, we introduce unMORE, a novel two-stage pipeline designed to identify many complex objects in real-world images. The key to our approach involves explicitly learning three levels of carefully defined object-centric representations in the first stage. Subsequently, our multi-object reasoning module utilizes these learned object priors to discover multiple objects in the second stage. Notably, this reasoning module is entirely network-free and does not require human labels. Extensive experiments demonstrate that unMORE significantly outperforms all existing unsupervised methods across 6 real-world benchmark datasets, including the challenging COCO dataset, achieving state-of-the-art object segmentation results. Remarkably, our method excels in crowded images where all baselines collapse.

Problem

• Unsupervised segmentation of multiple complex real-world objects
• Overcoming limitations of existing reconstruction-based methods
• Handling crowded images where baselines fail

Innovation

• Two-stage pipeline for unsupervised segmentation
• Learning object-centric representations in first stage
• Network-free reasoning module in second stage