notesum.ai
Published at November 27Online Knowledge Integration for 3D Semantic Mapping: A Survey
cs.RO
cs.CV
cs.LG
Released Date: November 27, 2024
Authors: Felix Igelbrink1, Marian Renz2, Martin Günther, Piper Powell2, Lennart Niecksch1, Oscar Lima1, Martin Atzmueller3, Joachim Hertzberg2
Aff.: 1German Research Center for Artificial Intelligence, Research Department Plan-based Robot Control, Osnabrück, Germany; 2Osnabrück University, Institute of Computer Science, Knowledge-Based Systems Group, Osnabrück, Germany; 3Osnabrück University, Institute of Computer Science, Semantic Information Systems Group, Osnabrück, Germany

| Name | Reference | Generation Method | Dataset | Application |
|---|---|---|---|---|
| 3-D scene graph | Kim et al. [69] | Image-level prediction using Factorizable Net | ScanNet [101] | Task planning, VQA |
| 3DSSG | Wald et al. [70] | Offline Prediction with GNN | Modified 3RScan [78] | Prediction |
| SceneGraph- Fusion | Wu et al. [79] | Online incremental prediction from single frames with GNN | 3DSSG [70], ScanNet [101] | Incremental Prediction |
| MonoSSG | Wu et al. [84] | Online incremental prediction from RGB sequences with GNN | 3DSSG | Incremental Prediction |
| KSGN | Qiu and Christensen [28] | Offline prediction of scene graphs with bridging to knowledge graphs | 3DSSG | Prediction |
| SGFormer | Lv et al. [86] | Offline prediction with transformer model and text embeddings | 3DSSG | Prediction |
| DGGN | Qi et al. [81] | Offline prediction on point clusters | s3DIS [102], 3DSSG, Paris-Lille-3D [103] | Prediction |
| LSSG+GSSG | Li et al. [80] | Online incremental prediction from single frames with GNN | AI2Thor [104] | Navigation |
| Object-based SLAM | Sünderhauf [92] | Construction of random pose- and landmark nodes | Custom indoor scenario | Object-centric navigation |
| SARP | Amiri et al. [95] | Predictions on 360-degree RGB images | Custom environment, AI2Thor | Object search |
| 3D VSG | Looper et al. [96] | GNN on ground truth scene graph | 3DSSG, 3RScan | Change detection |
| - | Kabalar et al. [71] | Online construction from single images | RIO10[105] | Visual localization |
| D-SCG | Giuliari et al. [89] | Offline construction from ground truth segmentations | ScanNet, ConceptNet [57] | Object localization in partial scenes |
| KiSGPM | Zhang et al. [44] | Offline prediction with knowledge-based meta-embeddings | 3DSSG | Prediction |
| MP-DGCNN + ENA-GNN | Han et al. [90] | Offline prediction with unbiased knowledge-based meta-embeddings | 3DSSG | Prediction |