AI Paper Atlas

Research map / 3D vision / Gaussian splatting

LightSplat: Fast and Memory-Efficient Open-Vocabulary 3D Scene Understanding in Five Seconds

Jaehun Bang, Jinhyeok Kim, Minji Kim, Seungheon Jeong, Kyungdon Joo

CVPR 2026 · 3D vision › Gaussian splatting · approach: Sparse-view Gaussian splatting for novel views · 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 3D-aware mask filtering (ablation)LightSplat · LERF-OVSmIoU: 29.31 -> 47.58Improved
add 3D-aware mask filtering (ablation)LightSplat · LERF-OVSmAcc@0.25: 25.96 -> 68.32Improved
add semantic-aware clustering (ablation)LightSplat · LERF-OVSmIoU: 19.56 -> 47.58Improved
add semantic-aware clustering (ablation)LightSplat · LERF-OVSmAcc@0.25: 18.97 -> 68.32Improved
add geometry-aware clustering (ablation)LightSplat · LERF-OVSmIoU: 2.01 -> 47.58Improved
add geometry-aware clustering (ablation)LightSplat · LERF-OVSmAcc@0.25: 2.27 -> 68.32Improved
add the ablated components (filtering and clustering) to the pipeline (ablation)LightSplat · LERF-OVSfeature distillation time: 4.42 s -> 4.55 sNo clear effect
replace OpenGaussian-style iterative per-Gaussian features with index-feature mapping and clusteringLightSplat · LERF-OVSmIoU: 42.15 -> 47.58Improved
replace OpenGaussian-style iterative per-Gaussian features with index-feature mapping and clusteringLightSplat · ScanNet (19 classes)mIoU: 29.43 -> 37.11Improved
replace LUDVIG per-Gaussian 2048-byte features with 2-byte indicesLightSplat · ScanNetmemory per Gaussian (bytes): 2048 -> 2Improved
replace LUDVIG iterative aggregation with single-step clusteringLightSplat · DL3DV-OVSmIoU: 29.21 -> 44.98Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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