Reproduce “EditSplat: Multi-View Fusion and Attention-Guided Optimization for View-Consistent 3D Scene Editing with 3D Gaussian Splatting” 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
Dong In Lee, Hyeongcheol Park, Jiyoung Seo, Eunbyung Park, Hyunje Park, Ha Dam Baek, Sangheon Shin, Sangmin Kim and 1 more
EditSplat edits 3D scenes from text prompts by combining neighboring edited views and feeding fused information back to a 2D editor. It uses diffusion cross-attention to guide Gaussian pruning and selective optimization. In 8 scenes with 2 prompts, it scores 0.1431 in CLIP text-image directional similarity, versus DGE’s 0.1222; this is an alignment proxy, not a measure of cross-view consistency.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.