Research map / 3D vision / 3D generation
3D-to-2D Distillation for Indoor Scene Parsing
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 3D-to-2D distillation to PSPNet-50 | PSPNet-50 · ScanNet-v2 validation | mIoU: 53.40 -> 58.22 | Helped |
| use simulated 3D features instead of predicting a depth map | PSPNet-50 · ScanNet-v2 validation | mIoU: 54.08 -> 58.22 | Helped |
| add 3D-to-2D distillation | HRNet-W48 + OCR · ScanNet-v2 validation | mIoU: 60.56 -> 61.36 | Helped |
| add 3D-to-2D distillation | PSPNet-50 · S3DIS | mIoU: 43.65 -> 46.42 | Helped |
| train with unpaired ScanNet-v2 3D data using adversarial distillation | PSPNet-50 · NYU-v2 20 class | mIoU: 49.50 to 51.70 | Helped |
| add 3D-to-2D distillation | PSPNet-50 trained on ScanNet-v2 · NYU-v2 20-class unseen domain | mIoU: 19.08 -> 27.22 | Helped |
| replace batch normalization with dimension normalization (ablation) | 3D-to-2D Distillation · ScanNet-v2 validation | mIoU: 57.64 -> 58.22 | Helped |
| replace batch normalization with dimension normalization (ablation) | 3D-to-2D Distillation · S3DIS | mIoU: 45.57 -> 46.42 | Helped |
| replace batch normalization with dimension normalization (ablation) | 3D-to-2D Distillation · NYU-v2 trained on ScanNet-v2 | mIoU: 26.54 -> 27.22 | Helped |
| use category-specific discriminators instead of a shared discriminator (ablation) | unpaired 3D-to-2D Distillation · NYU-v2 unpaired 2D-3D data | mIoU: 50.86 -> 51.84 | Helped |
| add 3D-to-2D distillation | PSPNet-50 · ScanNet-v2 depth reconstruction from res-block-4 | SSIM: baseline: 0.38 vs our: 0.47 | Helped |
Closest papers by meaning
- ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes (CVPR 2017 · similarity 0.89)
- Geometry-Aware Distillation for Indoor Semantic Segmentation (CVPR 2019 · similarity 0.88)
- Deep Supervision With Shape Concepts for Occlusion-Aware 3D Object Parsing (CVPR 2017 · similarity 0.88)
- Learning 3D Semantic Scene Graphs From 3D Indoor Reconstructions (CVPR 2020 · similarity 0.88)
- DistillNeRF: Perceiving 3D Scenes from Single-Glance Images by Distilling Neural Fields and Foundation Model Features (NeurIPS 2024 · similarity 0.86)
- SAM-Guided Masked Token Prediction for 3D Scene Understanding (NeurIPS 2024 · similarity 0.85)
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.