AI Paper Atlas

Research map / 3D vision / 3D generation

Towards 3D Object-Centric Feature Learning for Semantic Scene Completion

Weihua Wang, Yubo Cui, Xiangru Lin, Zhiheng Li, Zheng Fang

AAAI 2026 · 3D vision › 3D generation · approach: Semantic scene completion · 0 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 MobileSAM-guided SGA3D object-level attention (ablation)baseline SSC model · SemanticKITTI valmIoU: 15.80 -> 16.49Improved
add Global Similarity-Guided Attention to SGA3D (ablation)Ocean · SemanticKITTI valmIoU: 16.49 -> 16.80Improved
add local attention refinement of BEV features (ablation)Ocean · SemanticKITTI valmIoU: mIoU 16.80 -> 17.10, IoU 46.01 -> 45.39Mixed
add Dynamic Instance Decoder (ablation)Ocean · SemanticKITTI valmIoU: 17.10 -> 17.39Improved
add multi-scale image features to SGA3D (ablation)SGA3D · SemanticKITTI valmIoU: 16.64 -> 17.39Improved
add 3D depth-similarity extension to SGA (ablation)SGA3D · SemanticKITTI valIoU: 45.84 -> 46.40Improved
use SGA3D instead of DFA2D or DFA3D (ablation)Ocean · SemanticKITTI valmIoU: 16.39 (DFA2D), 16.97 (DFA3D) -> 17.39Improved
add GSGA module (ablation)Ocean · SemanticKITTI valmIoU: 16.98 -> 17.39Improved
use similarity-guided global features in GSGA instead of naive global operation (ablation)GSGA · SemanticKITTI valmIoU: 16.36 -> 17.39Improved
use dynamic instance selection (Gumbel-Softmax) instead of direct sum fusion (ablation)Dynamic Instance Decoder · SemanticKITTI valmIoU: 16.45 -> 17.39Improved
use dynamic selection instead of sigmoid weighting (ablation)Dynamic Instance Decoder · SemanticKITTI valIoU: 46.30 -> 46.40Improved
use Ocean instead of CGFormer on SSCBench-KITTI360Ocean · SSCBench-KITTI360 testmIoU: 20.05 -> 20.28Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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