AI Paper Atlas

Research map / 3D vision / Depth and stereo

Estimating Spatially-Varying Lighting in Urban Scenes with Disentangled Representation

Jiajun Tang, Yongjie Zhu, Haoyu Wang, Jun Hoong Chan, Si Li, Boxin Shi

ECCV 2022 · 3D vision › Depth and stereo · approach: Neural light field and illumination imaging · 19 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add panoramic cosine mask Mcos (ablation)local encoder-renderer · reconstructionMAE: 0.034 -> 0.028Improved
add cross rendering loss Lcr (ablation)local encoder-renderer · cross renderingMAE: 0.048 -> 0.031Improved
add local identity loss Lid (ablation)local encoder-renderer · cross renderingRMSE: 0.089 -> 0.079Improved
add cross rendering loss Lcr (ablation)local encoder-renderer · reconstructionMAE: 0.029 -> 0.028No clear effect
add sun visibility mask Mvis (ablation)local lighting estimator · local panoramaMAE: 0.140 -> 0.128Improved
add sun visibility mask Mvis (ablation)local lighting estimator · local relightingRMSE: 0.159 -> 0.145Improved
use disentangled representation instead of 5th-order SHlocal lighting estimator · local panoramaMAE: 0.190 -> 0.128Improved
use disentangled representation instead of 24 SG lobeslocal lighting estimator · local relightingMAE: 0.093 -> 0.075Improved
use SH parametrization instead of disentangled representationlocal lighting estimator · local panoramaRMSE: 2.394 -> 1.943Mixed
compare against SkyNet for relightingglobal lighting estimator · global relightingMAE: 0.226 -> 0.098Improved
compare against SOLID-Net for global panoramaglobal lighting estimator · global panoramaMAE: 0.384 -> 0.439Hurt
compare against SOLID-Net local lightinglocal lighting estimator · local relightingMAE: 0.186 -> 0.075Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy