XScale- NVS: Cross-Scale Novel View Synthesis with Hash Featurized Manifold

We propose XScale-NVS for high-fidelity cross-scale novel view synthesis of real-world large-scale scenes. Ex-isting representations based on explicit surface suffer from discretization resolution or <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$UV$</tex> distortion, while implicit volumetric representations lack scalability for large scenes due to the dispersed weight distribution and surface ambi-guity. In light of the above challenges, we introduce hash featurized manifold, a novel hash-based featurization cou-pled with a deferred neural rendering framework. This approach fully unlocks the expressivity of the representation by explicitly concentrating the hash entries on the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$2D$</tex> manifold, thus effectively representing highly detailed contents independent of the discretization resolution. We also in-troduce a novel dataset, namely <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$Gig{a}NVS$</tex>, to benchmark cross-scale, high-resolution novel view synthesis of real-world large-scale scenes. Our method significantly outper-forms competing baselines on various real-world scenes, yielding an average LPIPS that is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\sim$</tex> 40% lower than prior state-of-the-art on the challenging <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$Gig{a}NVS$</tex> benchmark. Please see our project page at: xscalenvs.github.io.