Research map / Point clouds / Segmentation and detection
Semantic Segmentation for Real Point Cloud Scenes via Bilateral Augmentation and Adaptive Fusion
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 geometric bilateral offsets and geometric augmentation loss (ablation) | Bilateral Context Block · S3DIS Area 5 | mIoU: 61.8 -> 64.6 | Helped |
| Add semantic augmentation loss (ablation) | Bilateral Context Block · S3DIS Area 5 | mIoU: 64.2 -> 64.3 | No clear effect |
| Use mixed max-and-mean local aggregation (ablation) | Bilateral Context Block · S3DIS Area 5 | mIoU: 64.6 -> 65.4 | Helped |
| Use only the final-resolution feature map instead of multi-resolution maps (ablation) | Adaptive Fusion Module · S3DIS Area 5 | mIoU: 65.4 -> 64.1 | Hurt |
| Fuse multi-resolution maps by summation without learned weights (ablation) | Adaptive Fusion Module · S3DIS Area 5 | mIoU: 65.4 -> 64.7 | Hurt |
| Fuse multi-resolution maps by element-wise multiplication (ablation) | Adaptive Fusion Module · S3DIS Area 5 | mIoU: 65.4 -> 64.2 | Hurt |
| Concatenate multi-resolution maps instead of using point-level adaptive fusion (ablation) | Adaptive Fusion Module · S3DIS Area 5 | mIoU: 65.4 -> 65.1 | Hurt |
| Use one scalar per resolution instead of point-level fusion weights (ablation) | Adaptive Fusion Module · S3DIS Area 5 | mIoU: 65.4 -> 65.1 | Hurt |
| Replace farthest point sampling with random sampling (ablation) | BAAF-Net · S3DIS Area 5 | mIoU: 65.4 -> 64.8 | Hurt |
| Use dilated k-nearest-neighbor search with factor 2 (ablation) | BAAF-Net · S3DIS Area 5 | mIoU: 65.4 -> 62.5 | Hurt |
| Set equal weights of 0.3 for all augmentation losses (ablation) | BAAF-Net · S3DIS Area 5 | mIoU: 65.4 -> 64.0 | Hurt |
| Use four resolutions instead of five (ablation) | BAAF-Net · S3DIS Area 5 | mIoU: 65.4 -> 63.5 | Hurt |
Closest papers by meaning
- Point-MoE: Large-Scale Multi-Dataset Training with Mixture-of-Experts for 3D Semantic Segmentation (ICLR 2026 · similarity 0.91)
- Associatively Segmenting Instances and Semantics in Point Clouds (CVPR 2019 · similarity 0.91)
- PointGS: Semantic-Consistent Unsupervised 3D Point Cloud Segmentation with 3D Gaussian Splatting (CVPR 2026 · similarity 0.90)
- Exploiting Local and Global Structure for Point Cloud Semantic Segmentation with Contextual Point Representations (NeurIPS 2019 · similarity 0.90)
- Geometrically-driven Aggregation for Zero-shot 3D Point Cloud Understanding (CVPR 2024 · similarity 0.90)
- Multi-Modality Affinity Inference for Weakly Supervised 3D Semantic Segmentation (AAAI 2024 · similarity 0.88)
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.