Reproduce “MAVE: Reconstructing Editable 3D Blockouts from Film Shots” 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
Yifan Chang, Jian Yang, Jiangnan Shao, Xiaofeng Wang, Zhaoxin Fan, Renjing Xu, Ming Lu, Xingang Wang and 1 more
MAVE turns film shots into editable 3D blockouts by measuring camera motion, relative depth, and human motion before an agent builds a Blender scene and checks edits against evidence. On MAVEBench, DeepSeek-1017 with MAVE scores 73.2 versus 37.2 for the same agent without MAVE; this score includes recovered cameras and applies to short film shots with people.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.