Research map / Detection and segmentation / 3D detection and driving
Language-Assisted 3D Feature Learning for Semantic Scene Understanding
Junbo Zhang, Guofan Fan, Guanghan Wang, Zhengyuan Su, Kaisheng Ma, Li Yi
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 language-assisted training with object-level relationship and attribute supervision | VoteNet · ScanNetV2 object detection | mAP@0.5: 31.67 -> 33.41 | Improved |
| Add language-assisted training | MLCVNet · ScanNetV2 object detection | mAP@0.5: 30.18 -> 33.69 | Improved |
| Add language-assisted training | VoteNet · ScanNetV2 detection with 1 or 4 boxes per scene | mAP@0.25: 13.74 -> 18.55; 23.08 -> 30.61 | Improved |
| Add language-assisted training to PointGroup | PointGroup · ScanNetV2 instance segmentation | mAP@0.5: 55.7 -> 56.9 | Improved |
| Train the 3D backbone with language assistance | ScanRefer · ScanRefer val visual grounding, overall 3D input | Acc@0.5IoU: 25.81 -> 28.00 | Improved |
| Train the 3D backbone with language assistance | Scan2Cap · ScanRefer val captioning | CiDEr@0.25IoU: 50.95 -> 52.97 | Improved |
| Add relationship classification auxiliary supervision (ablation) | VoteNet · ScanNetV2 object detection | mAP@0.5: 31.67 -> 32.47 | Improved |
| Add attribute classification auxiliary supervision (ablation) | VoteNet · ScanNetV2 object detection | mAP@0.5: 31.67 -> 32.80 | Improved |
| Add both relationship and attribute classification tasks (ablation) | VoteNet · ScanNetV2 object detection | mAP@0.5: 31.67 -> 33.41 | Improved |
| Increase the number of relationship categories from 5 to 16 (ablation) | VoteNet with language-assisted training · ScanNetV2 object detection | mAP@0.5: 32.48 -> 33.46 | Improved |
| Increase the number of attribute categories from 13 to 37 (ablation) | VoteNet with language-assisted training · ScanNetV2 object detection | mAP@0.5: 32.81 -> 33.31 | Improved |
| Use binary cross-entropy rather than cross-entropy for relation classification (ablation) | VoteNet with language-assisted training · ScanNetV2 object detection | mAP@0.5: 32.51 -> 33.13 | Improved |
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.
- MLCVNet — MLCVNet: Multi-Level Context VoteNet for 3D Object Detection (CVPR 2020) · better on 1 of 1 shared result
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
- Lexicon3D: Probing Visual Foundation Models for Complex 3D Scene Understanding (NeurIPS 2024 · similarity 0.90)
- RegionPLC: Regional Point-Language Contrastive Learning for Open-World 3D Scene Understanding (CVPR 2024 · similarity 0.90)
- A Unified Framework for 3D Scene Understanding (NeurIPS 2024 · similarity 0.90)
- Masked Point-Entity Contrast for Open-Vocabulary 3D Scene Understanding (CVPR 2025 · similarity 0.89)
- MLLMs Need 3D-Aware Representation Supervision for Scene Understanding (NeurIPS 2025 · similarity 0.89)
- Segment Any 3D Object with Language (ICLR 2025 · similarity 0.89)
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.
