Reproduce “PVDepth: Panoramic Video Depth Estimation via Geometry-Aware Spatiotemporal Adaptation” 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
Chuanxin Song, Peixi Peng
PVDepth adapts a pretrained video-depth model to panoramic video by annealing spherical noise into training and processing temporal features in cubemap views before returning them to the panorama. On synthetic outdoor PanoCARLA, it reports average AbsRel 0.199 and δ₁ 0.798, beating a 40K-step fine-tuned baseline under the paper’s legacy protocol; depth is relative, not metric.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.