Reproduce “MLLM-GS: Multimodal Large Language Model Prior-Guided Fast Novel View Synthesis in Extreme Environments” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zhiliang Wu, Xinye Zheng, Kun Li, Fei Wang, Hao Guo, Yanyan Wei, Guanghui Yue, Beier Zhu and 2 more

MLLM-GS reconstructs novel views from degraded multi-view images by enhancing them with a frozen multimodal model, correcting outputs using three paired clean and degraded views per scene. It trains efficient 3D Gaussian Splatting. On two scenes, it reaches 27 dB PSNR on smoke and 24 dB on low light after 5,000 iterations, in 100 seconds; this does not establish robustness.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.