1 Literature Review

This section contains detailed reviews of key papers and methods related to Next-Best-View planning and foundation models.

1.1 Key Papers

1.1.1 Next-Best-View Planning

• VIN-NBV: Learning-based RRI prediction for quality-driven view planning
• GenNBV: Continuous action spaces and reinforcement learning approaches

1.1.2 Foundation Models & Scene Understanding

• EFM3D/EVL: Egocentric foundation models for 3D perception and voxel lifting
• SceneScript: Structured scene language for semantic scene manipulation

1.2 Local Corpus

The full source of every referenced paper is mirrored under literature/tex-src/. Key entry points:

• arXiv-EFM3D/main.tex
• arXiv-GenNBV/main.tex
• arXiv-VIN-NBV/main.tex
• arXiv-project-aria/main.tex
• arXiv-scene-script/main.tex

Each directory contains the canonical figures (figures/), tables, and supplemental PDFs required to reproduce results or cite specific numbers in our documentation.

1.3 Research Context

These literature reviews form the theoretical foundation for integrating pre-trained egocentric models with quality-driven NBV planning. The combination of VIN-NBV’s RRI optimization with EFM3D’s foundation model capabilities enables better generalization to complex indoor environments compared to traditional coverage-based approaches.

1.4 Navigation

Return to main documentation or explore specific metrics and evaluation approaches.