Research map / Point clouds / Shape and surface completion
Unsupervised 3D Learning for Shape Analysis via Multiresolution Instance Discrimination
Read the paperCheck 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 |
|---|---|---|---|
| Fine-tune MID-Net pre-trained with MID instead of training from scratch | MID-FC · ModelNet40 | accuracy: 92.9 -> 93.0 | Helped |
| Fine-tune MID-Net pre-trained with MID instead of training from scratch | MID-FC · PartNet | mIoU: 58.4 -> 60.8 | Helped |
| Fine-tune MID-Net pre-trained with MID instead of training from scratch | MID-FC · ShapeNet Part | C.mIoU: 84.1 -> 84.3 | Helped |
| Add point-instance discrimination to shape-instance discrimination (ablation) | MID-Net · ModelNet40, 1% labeled data | accuracy: 60.5 -> 61.8 | Helped |
| Add point-instance discrimination to shape-instance discrimination (ablation) | MID-Net · ShapeNet Part, 1% labeled data | category IoU: 61.6 -> 74.6 | Helped |
| Add shape-instance discrimination to point-instance discrimination (ablation) | MID-Net · ModelNet40, 1% labeled data | accuracy: 54.4 -> 61.8 | Helped |
| Add shape-instance discrimination to point-instance discrimination (ablation) | MID-Net · ShapeNet Part, 1% labeled data | category IoU: 72.2 -> 74.6 | Helped |
| Retain the first HRNet fusion layer (ablation) | MID-Net · ModelNet40 | accuracy: 89.9 -> 90.3 | Helped |
| Retain all HRNet fusion layers (ablation) | MID-Net · ModelNet40 | accuracy: 88.2 -> 90.3 | Helped |
| Use HRNet rather than U-Net (ablation) | MID-Net · ModelNet40 | accuracy: 89.3 -> 90.3 | Helped |
| Use HRNet rather than U-Net (ablation) | MID-Net · ShapeNet Part | Seg.: 84.5 -> 85.5 | Helped |
| Retain rotation augmentation (ablation) | MID-Net · ModelNet40 | accuracy: 89.9 -> 90.3 | Helped |
Closest papers by meaning
- Unsupervised Learning of Intrinsic Structural Representation Points (CVPR 2020 · similarity 0.88)
- Symmetric Shape-Preserving Autoencoder for Unsupervised Real Scene Point Cloud Completion (CVPR 2023 · similarity 0.88)
- Learning a Structured Latent Space for Unsupervised Point Cloud Completion (CVPR 2022 · similarity 0.88)
- PartNet: A Recursive Part Decomposition Network for Fine-Grained and Hierarchical Shape Segmentation (CVPR 2019 · similarity 0.88)
- Unsupervised Deep Shape Descriptor With Point Distribution Learning (CVPR 2020 · similarity 0.88)
- Unsupervised Template-assisted Point Cloud Shape Correspondence Network (CVPR 2024 · similarity 0.86)
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.