AI Paper Atlas

Research map / 3D vision / Gaussian splatting

2D Gaussian Splatting-based Sparse-view Transparent Object Depth Reconstruction via Physics Simulation for Scene Update

Jeongyun Kim, Seunghoon Jeong, Giseop Kim, Myung-Hwan Jeon, Eunji Jun, Ayoung Kim

ICCV 2025 · 3D vision › Gaussian splatting · approach: Compact and efficient Gaussian splatting · 1 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 object-aware loss (ablation)TRAN-D · synthetic depth, t=0MAE: 0.0447 -> 0.0419Improved
add object-aware loss (ablation)TRAN-D · synthetic depth, t=0Gaussian count: 35983 -> 33482Improved
add object-aware loss (ablation)TRAN-D · synthetic depth, t=1MAE: 0.0932 -> 0.0886Improved
add physics simulation for scene update (ablation)TRAN-D · synthetic depth, t=1MAE: 0.0891 -> 0.0886Improved
add physics simulation for scene update (ablation)TRAN-D · synthetic depth, t=1RMSE: 0.1945 -> 0.1936Improved
change from 3 to 12 training views (ablation)TRAN-D · synthetic depth, t=0MAE: 0.0405 -> 0.0448No clear effect
change from 3 to 12 training views (ablation)InstantSplat · synthetic depth, t=0MAE: 0.1306 -> 0.2062Hurt
change from 3 to 12 training views (ablation)FSGS · synthetic depth, t=0MAE: 0.1846 -> 0.1426Improved
use TRAN-D object-only reconstruction vs best baselineTRAN-D vs TranSplat · TRansPose synthetic depth, t=0MAE: 0.0632 -> 0.0380Improved
use TRAN-D vs 2DGS after object removalTRAN-D vs 2DGS · TRansPose synthetic depth, t=1MAE: 0.0849 -> 0.0864Mixed
use TRAN-D with single bird-eye view for post-change update vs 6-image baselinesTRAN-D · efficiency, t=1total training time (s): 417.0 -> 13.8Improved
use TRAN-D object-only Gaussians vs 3DGSTRAN-D · efficiency, t=0Gaussian count: 175.2k -> 33.5kImproved

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