Reproduce “E-WAVE: Event-based Continuous Optical Flow via Warping-Aligned Visual Encoding” 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
Jiale Wu, Xiaoyang Bai, Haoming Yu, Yiwei Chen, Yifan Peng, Weiwei Xu
E-WAVE estimates continuous pixel motion paths from event signals by warping event and optional image features along Bézier trajectories and refining them with global attention. On synthetic MultiFlow, its event-RGB model achieves trajectory endpoint error (TEPE) of 1.25, versus 1.65 for STFlow†, while endpoint error is similar (2.04 versus 1.98). Real-world head-mounted tests are zero-shot and qualitative, without ground-truth motion.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.