AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Hierarchical Disentangled Representation Learning for Outdoor Illumination Estimation and Editing

Piaopiao Yu, Jie Guo, Fan Huang, Cheng Zhou, Hongwei Che, Xiao Ling, Yanwen Guo

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 18 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 information-theory loss L_info (ablation)HDSky AE1 · Sunny/50 panoramasRMSE: 0.785 -> 0.569Improved
add sun loss L_sun (ablation)HDSky AE1 · Sunny/50 panoramasRMSE: 0.757 -> 0.569Improved
add sun position map P_pos (ablation)HDSky AE1 · Sunny/50 panoramasRMSE: 0.774 -> 0.569Improved
add information-theory loss L_info (ablation)HDSky AE1 · Sunny/50 rendersRMSE: 0.028 -> 0.025Improved
add sun loss L_sun (ablation)HDSky AE1 · Sunny/50 rendersRMSE: 0.033 -> 0.025Improved
add sun position map P_pos (ablation)HDSky AE1 · Sunny/50 rendersRMSE: 0.026 -> 0.025Improved
use sky:sun latent dimension ratio 8:53 instead of 16:45 (ablation)HDSky predictor · Sunny images (300)RMSE: 11.84 -> 10.28Improved
use ratio 8:53 instead of 24:37 (ablation)HDSky predictor · Sunny images (300)RMSE: 12.31 -> 10.28Improved
use ratio 8:53 instead of 32:29 (ablation)HDSky predictor · Sunny images (300)RMSE: 14.18 -> 10.28Improved
use ratio 8:53 instead of 48:13 (ablation)HDSky predictor · Sunny images (300)RMSE: 10.51 -> 10.28Improved
disentangled sky/sun representation instead of single latent vector (SkyNet)HDSky reconstruction · Sunny/50 panoramasRMSE: 1.532 -> 0.569Improved
disentangled representation instead of SkyNet single vectorHDSky predictor · Cloudy/100 rendersRMSE: 0.101 -> 0.030Improved

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