Research map / 3D vision / 3D generation
ReferIt3D: Neural Listeners for Fine-Grained 3D Object Identification in Real-World Scenes
Panos Achlioptas, Ahmed Abdelreheem, Fei Xia, Mohamed Elhoseiny, Leonidas Guibas
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add auxiliary classification losses (ablation) | ReferIt3DNet · Nr3D test | listening accuracy: 32.4% -> 35.6% | Improved |
| add auxiliary classification losses (ablation) | ReferIt3DNet · Sr3D test | listening accuracy: 38.4% -> 39.8% | Improved |
| add auxiliary classification losses (ablation) | V + L + C · Sr3D test | listening accuracy: 34.7% -> 37.2% | Improved |
| add auxiliary classification losses (ablation) | V + L · Nr3D test | listening accuracy: 26.12% -> 26.62% | Improved |
| add graph (DGCN) scene context to V + L + C (ablation) | V + L + C · Nr3D test (with aux losses) | listening accuracy: 28.51% -> 35.6% | Improved |
| add holistic PointNet++ scene context (ablation) | V + L · Nr3D test (with aux losses) | listening accuracy: 26.62% -> 28.51% | Improved |
| fuse language both before and after DGCN instead of only before (ablation) | ReferIt3DNet · Nr3D test (with aux losses) | listening accuracy: 33.0% -> 35.6% | Improved |
| fuse language both before and after DGCN instead of only after (ablation) | ReferIt3DNet · Nr3D test (with aux losses) | listening accuracy: 33.4% -> 35.6% | Improved |
| fuse language both places instead of only after DGCN (ablation) | ReferIt3DNet · Sr3D test (with aux losses) | listening accuracy: 41.0% -> 39.8% | Hurt |
| add Sr3D training data | ReferIt3DNet · Nr3D test (overall) | listening accuracy: 35.6% -> 37.2% | Improved |
| add Sr3D+ training data | ReferIt3DNet · Nr3D test (overall) | listening accuracy: 35.6% -> 37.6% | Improved |
| add Sr3D+ training data | ScanRefer · ScanRefer (Overall) | P@0.5: 16.90% -> 20.12% | Improved |
Closest papers by meaning
- Chat-Scene: Bridging 3D Scene and Large Language Models with Object Identifiers (NeurIPS 2024 · similarity 0.89)
- MMScan: A Multi-Modal 3D Scene Dataset with Hierarchical Grounded Language Annotations (NeurIPS 2024 · similarity 0.88)
- SceneFun3D: Fine-Grained Functionality and Affordance Understanding in 3D Scenes (CVPR 2024 · similarity 0.88)
- OpenLex3D: A Tiered Benchmark for Open-Vocabulary 3D Scene Representations (NeurIPS 2025 · similarity 0.88)
- 3D-PLOT-LLM: Part-Level Object Tokens for 3D Large Language Models (NeurIPS 2026 · similarity 0.87)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.87)
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.
