Research map / 3D vision / 3D generation
Multiview Compressive Coding for 3D Reconstruction
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 |
|---|---|---|---|
| Use MCC instead of NerFormer with depth input and supervision | CO3D-v2 reconstruction · CO3D-v2 unseen categories | Abs: 2.28 -> 1.17 | Helped |
| Use MCC instead of NerFormer with depth input and supervision | CO3D-v2 reconstruction · CO3D-v2 unseen categories | MSE: 24.1 -> 13.6 | Helped |
| Use MCC rather than DRDF with its original method | Hypersim scene reconstruction · Hypersim | F1: 2.0 -> 2.8 | Helped |
| Use separate RGB and geometry encoders rather than a shared encoder (ablation) | MCC · CO3D-v2 held-out categories | F1: 52.5 -> 56.7 | Helped |
| Use a transformer rather than an MLP for geometry encoding (ablation) | MCC · CO3D-v2 held-out categories | F1: 54.5 -> 56.7 | Helped |
| Use a transformer rather than PointNet for geometry encoding (ablation) | MCC · CO3D-v2 held-out categories | F1: 56.6 -> 56.7 | No clear effect |
| Use uniform rather than contrastive-style query sampling (ablation) | MCC · CO3D-v2 held-out categories | F1: 55.6 -> 56.7 | No clear effect |
| Keep detailed input tokens instead of globally pooled conditioning (ablation) | MCC · CO3D-v2 held-out categories | F1: 54.5 -> 56.7 | Helped |
| Keep detailed input tokens instead of location-pooled conditioning (ablation) | MCC · CO3D-v2 held-out categories | F1: 28.2 -> 56.7 | Helped |
| Use concatenated query and input tokens with attention instead of cross-attention (ablation) | MCC decoder · CO3D-v2 held-out categories | F1: 43.7 -> 56.7 | Helped |
| Use the query-token attention decoder rather than location features and an MLP (ablation) | MCC decoder · CO3D-v2 held-out categories | F1: 28.2 -> 56.7 | Helped |
| Add RGB image encoding to geometry encoding (ablation) | MCC · CO3D-v2 held-out categories | F1: 53.9 -> 56.7 | Helped |
Closest papers by meaning
- NU-MCC: Multiview Compressive Coding with Neighborhood Decoder and Repulsive UDF (NeurIPS 2023 · similarity 0.91)
- What Do Single-View 3D Reconstruction Networks Learn? (CVPR 2019 · similarity 0.89)
- Slice3D: Multi-Slice Occlusion-Revealing Single View 3D Reconstruction (CVPR 2024 · similarity 0.88)
- Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis (NeurIPS 2026 · similarity 0.87)
- Sparse3D: Distilling Multiview-Consistent Diffusion for Object Reconstruction from Sparse Views (AAAI 2024 · similarity 0.87)
- Duoduo CLIP: Efficient 3D Understanding with Multi-View Images (ICLR 2025 · similarity 0.87)
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.