Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

MMGDreamer: Mixed-Modality Graph for Geometry-Controllable 3D Indoor Scene Generation

Authors: Zhifei Yang, Keyang Lu, Chao Zhang, Jiaxing Qi, Hanqi Jiang, Ruifei Ma, Shenglin Yin, Yifan Xu, Mingzhe Xing, Zhen Xiao, Jieyi Long, Xiangde Liu, Guangyao Zhai

AAAI 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experimental results demonstrate that MMGDreamer exhibits superior control of object geometry, achieving state-of-the-art scene generation performance. Extensive experiments on the SG-FRONT dataset demonstrate that MMGDreamer attains higher fidelity and geometric controllability, and achieves state-of-the-art performance in scene synthesis, outperforming existing methods by a large margin.
Researcher Affiliation Collaboration 1 School of Computer Science, Peking University 2 School of Artificial Intelligence, Beihang University 3 Beijing Digital Native Digital City Research Center 4 School of Computer Science and Engineering, Beihang University 5 Theta Labs, Inc. 6 Technical University of Munich
Pseudocode No The paper describes methods using mathematical formulas and architectural descriptions of modules but does not contain explicit pseudocode or algorithm blocks.
Open Source Code No Project page https://yangzhifeio.github.io/project/MMGDreamer. The paper provides a project page link, which is typically a demonstration or overview page, but does not explicitly state that the source code for the methodology is openly available or provide a direct link to a code repository.
Open Datasets Yes We validate our approach using the SG-FRONT dataset (Zhai et al. 2024c), which provides comprehensive scene-graph annotations for indoor scenes.
Dataset Splits No The paper mentions using the SG-FRONT dataset and extracted images from 3D-FUTURE, but it does not specify any training/test/validation split percentages or counts. It states: "We then applied a random mask to mask the text, images, and relationships between nodes in the Full-Modality Graph, producing the Mixed-Modality Graph."
Hardware Specification Yes All experiments are performed on a single NVIDIA A100 GPU with 80 GB memory.
Software Dependencies No The paper mentions using the Adam W optimizer but does not provide specific version numbers for any programming languages, libraries, or other software dependencies.
Experiment Setup Yes We train our models using the Adam W optimizer, initializing the learning rate at 1 10 4 and utilizing a batch size of 128. The weighting factors for our loss components, α1 and α2, are consistently set to 1.0.