Reproduce “Beyond Row Alignment: Virtual-Camera-Aware Online Stereo Rectification” 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
Yizhao Peng, Rui Gong, Zaiwang Gu, Xudong Jiang, Jun Cheng
VirtualRect aligns stereo image rows, supervises the full-image warp, and supplies the matching virtual camera for depth. It estimates vertical-flow error using pretrained RAFT, rather than ground-truth correspondences. On Carla-Flowguided, error is 0.43 pixels versus RsCeR’s 0.52, a 17% reduction. Carla depth MAE falls 30–41% across three matchers. Results assume known intrinsics and bounded camera changes; real-world measurement precision remains limited.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.