Reproduce “WayveScenes101: A Dataset and Benchmark for Novel View Synthesis in Autonomous Driving” 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

Jannik Zürn, Paul Gladkov, Sofía Dudas, Fergal Cotter, Sofi Toteva, Jamie Shotton, Vasiliki Simaiaki, Nikhil Mohan

WayveScenes101 introduces a new dataset and benchmark specifically for novel view synthesis within autonomous driving environments. It comprises 101 diverse driving scenes, yielding 101,000 high-resolution images with accurate camera poses and comprehensive metadata, alongside a dedicated evaluation protocol for off-axis generalization.

Key insights

  • Applying novel view synthesis to autonomous driving requires datasets that capture complex dynamics, challenging photometric variations, and constrained camera viewpoints.
  • A dedicated evaluation strategy, focusing on 'off-axis reconstruction quality,' is essential for NVS models to demonstrate true 3D understanding and generalization for AD applications.
  • Rich, granular scene metadata is critical for enabling detailed analysis of NVS model performance under specific environmental and traffic conditions.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.