Research map / Point clouds / Shape and surface completion
Denoise and Contrast for Category Agnostic Shape Completion
Antonio Alliegro, Diego Valsesia, Giulia Fracastoro, Enrico Magli, Tatiana Tommasi
Check 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 denoising pretraining (ablation) | DeCo · Known categories | Chamfer Distance ×10^4: 23.865 -> 21.022 | Improved |
| add contrastive pretraining (ablation) | DeCo · Known categories | Chamfer Distance ×10^4: 23.865 -> 23.067 | Improved |
| combine denoising and contrastive pretraining (ablation) | DeCo · Known categories | Chamfer Distance ×10^4: 21.022 -> 18.742 | Improved |
| add frame reconstruction constraint (ablation) | DeCo with denoising and contrastive pretraining · Known categories | Chamfer Distance ×10^4: 18.742 -> 16.517 | Improved |
| replace contrastive pretraining with supervised classification (ablation) | DeCo · Known categories | Chamfer Distance ×10^4: 16.517 -> 18.699 | Hurt |
| replace the GPDNet denoising encoder with DGCNN (ablation) | DeCo · Known categories | Chamfer Distance ×10^4: 16.517 -> 19.667 | Hurt |
| use contrastive pairs instead of quadruplets (ablation) | DeCo global encoder · Known categories | Chamfer Distance ×10^4: 16.517 -> 18.030 | Hurt |
| vary the auxiliary frame size and decoder node count (ablation) | DeCo · Known categories | Chamfer Distance ×10^4: 16.517; 17.693; 18.068; 18.129; 18.595; 19.001 | Mixed |
| train without denoising, contrastive pretraining, or frame constraint (ablation) | DeCo · Novel categories, similar | Chamfer Distance ×10^4: 19.364 -> 26.811 | Hurt |
| train without denoising, contrastive pretraining, or frame constraint (ablation) | DeCo · Novel categories, dissimilar | Chamfer Distance ×10^4: 32.945 -> 40.419 | Hurt |
| add denoising and contrastive pretraining (ablation) | DeCo · Novel categories, similar | Chamfer Distance ×10^4: 26.811 -> 19.364 | Improved |
| add denoising and contrastive pretraining (ablation) | DeCo · Novel categories, dissimilar | Chamfer Distance ×10^4: 40.419 -> 32.945 | 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.
- PF-Net — PF-Net: Point Fractal Network for 3D Point Cloud Completion (CVPR 2020) · better on 5 of 5 shared results
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
- Complete Structure Guided Point Cloud Completion via Cluster- and Instance-Level Contrastive Learning (NeurIPS 2025 · similarity 0.90)
- Cascaded Refinement Network for Point Cloud Completion with Self-supervision (TPAMI 2021 · similarity 0.90)
- Diverse Shape Completion via Style Modulated Generative Adversarial Networks (NeurIPS 2023 · similarity 0.89)
- Multimodal Shape Completion via Conditional Generative Adversarial Networks (ECCV 2020 · similarity 0.89)
- Learning 3D Shape Completion From Laser Scan Data With Weak Supervision (CVPR 2018 · similarity 0.89)
- Learning 3D Shape Completion Under Weak Supervision (IJCV 2018 · 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.
