SGR3 Model: Scene Graph Retrieval-Reasoning Model in 3D
This work addresses the problem of generating 3D scene graphs for robots and humans without requiring multi-modal data or heuristic graph construction, offering an alternative to existing GNN-based methods.
The SGR3 Model is a training-free framework that uses multi-modal large language models (MLLMs) with retrieval-augmented generation (RAG) to generate 3D semantic scene graphs. It avoids explicit 3D reconstruction and enhances relational reasoning by retrieving semantically aligned scene graphs, achieving competitive performance compared to training-free baselines and GNN-based expert models.
3D scene graphs provide a structured representation of object entities and their relationships, enabling high-level interpretation and reasoning for robots while remaining intuitively understandable to humans. Existing approaches for 3D scene graph generation typically combine scene reconstruction with graph neural networks (GNNs). However, such pipelines require multi-modal data that may not always be available, and their reliance on heuristic graph construction can constrain the prediction of relationship triplets. In this work, we introduce a Scene Graph Retrieval-Reasoning Model in 3D (SGR3 Model), a training-free framework that leverages multi-modal large language models (MLLMs) with retrieval-augmented generation (RAG) for semantic scene graph generation. SGR3 Model bypasses the need for explicit 3D reconstruction. Instead, it enhances relational reasoning by incorporating semantically aligned scene graphs retrieved via a ColPali-style cross-modal framework. To improve retrieval robustness, we further introduce a weighted patch-level similarity selection mechanism that mitigates the negative impact of blurry or semantically uninformative regions. Experiments demonstrate that SGR3 Model achieves competitive performance compared to training-free baselines and on par with GNN-based expert models. Moreover, an ablation study on the retrieval module and knowledge base scale reveals that retrieved external information is explicitly integrated into the token generation process, rather than being implicitly internalized through abstraction.