RelGraphOV constructs relationship-aware 3D scene graphs via VLM reasoning and uses a dual-stream gated GAT to refine open-vocabulary 3D segmentation, achieving state-of-the-art on ScanNetV2, ScanNet200, ScanNet++, and Replica.
In: Inter- national Conference on 3D Vision, 3DV 2024, Davos, Switzerland, March 18-21
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Beyond Isolated Objects: Relationship-aware Open Vocabulary Scene Understanding via 3D Scene Graph Analysis
RelGraphOV constructs relationship-aware 3D scene graphs via VLM reasoning and uses a dual-stream gated GAT to refine open-vocabulary 3D segmentation, achieving state-of-the-art on ScanNetV2, ScanNet200, ScanNet++, and Replica.