Reproduce “Many Eyes, One World: Feed-Forward 3D Reconstruction from Mixed Cameras” 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
Qiaoge Li, Yifan Zhan, Haijun Yang, Haiyang Liu, Yiyi Cai, Chenchi Luo
MEOW adapts a perspective-pretrained reconstruction network to mixed-camera images using synthetic tuples with varied projections, full-field-of-view resizing, and aspect-ratio embeddings. It predicts a shared 3D reconstruction in one pass without camera labels or calibration. Its scores are 80.4 versus 54.3 mAA@30 on 2D3DS and 79.4 versus 29.3 AUC@30 on a four-view laser track, but it trails on 15–24-view cases.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.