Learning Physical Graph Representations from Visual Scenes

TL;DR

This paper introduces Physical Scene Graphs (PSGs) and PSGNet, a novel neural network architecture that explicitly encodes objects, parts, and their physical properties in scenes, improving scene understanding beyond traditional CNNs.

Contribution

The paper proposes PSGs as hierarchical graph representations of scenes and PSGNet to learn these structures, integrating feedback, graph pooling, and perceptual grouping for enhanced scene segmentation.

Findings

PSGNet outperforms existing self-supervised methods on scene segmentation.

PSGNet generalizes well to unseen objects and arrangements.

Learned latent attributes capture intuitive scene properties.

Abstract

Convolutional Neural Networks (CNNs) have proved exceptional at learning representations for visual object categorization. However, CNNs do not explicitly encode objects, parts, and their physical properties, which has limited CNNs' success on tasks that require structured understanding of visual scenes. To overcome these limitations, we introduce the idea of Physical Scene Graphs (PSGs), which represent scenes as hierarchical graphs, with nodes in the hierarchy corresponding intuitively to object parts at different scales, and edges to physical connections between parts. Bound to each node is a vector of latent attributes that intuitively represent object properties such as surface shape and texture. We also describe PSGNet, a network architecture that learns to extract PSGs by reconstructing scenes through a PSG-structured bottleneck. PSGNet augments standard CNNs by including: recurrent feedback connections to combine low and high-level image information; graph pooling and vectorization operations that convert spatially-uniform feature maps into object-centric graph structures; and perceptual grouping principles to encourage the identification of meaningful scene elements. We show that PSGNet outperforms alternative self-supervised scene representation algorithms at scene segmentation tasks, especially on complex real-world images, and generalizes well to unseen object types and scene arrangements. PSGNet is also able learn from physical motion, enhancing scene estimates even for static images. We present a series of ablation studies illustrating the importance of each component of the PSGNet architecture, analyses showing that learned latent attributes capture intuitive scene properties, and illustrate the use of PSGs for compositional scene inference.