Research map / 3D vision / Face and shape
Few-Shot Single-View 3D Reconstruction with Memory Prior Contrastive Network
Zhen Xing, Yijiang Chen, Zhixin Ling, Xiangdong Zhou, Yu Xiang
Read the paperCheck 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 |
|---|---|---|---|
| use multi-head attention over top-5 retrieved shapes instead of only the top-1 retrieved shape (ablation) | MPCN · ShapeNet / Pascal3D+ (5-shot) | mean IoU (%): 39.95 -> 47.53 (ShapeNet); 31.65 -> 33.54 (Pascal3D+) | Improved |
| use attention prior module instead of LSTM fusion (ablation) | MPCN · ShapeNet / Pascal3D+ (5-shot) | mean IoU (%): 40.52 -> 47.53 (ShapeNet); 25.63 -> 33.54 (Pascal3D+) | Improved |
| use retrieved shape priors instead of random prior vectors (ablation) | MPCN · ShapeNet / Pascal3D+ (5-shot) | mean IoU (%): 35.52 -> 47.53 (ShapeNet); 22.05 -> 33.54 (Pascal3D+) | Improved |
| use attention fusion instead of averaging the top-5 retrieved features (ablation) | MPCN · ShapeNet / Pascal3D+ (5-shot) | mean IoU (%): 41.32 -> 47.53 (ShapeNet); 29.85 -> 33.54 (Pascal3D+) | Improved |
| use 3D-aware contrastive loss instead of InfoNCE (ablation) | MPCN · ShapeNet / Pascal3D+ (5-shot) | mean IoU (%): 42.98 -> 47.53 (ShapeNet); 27.98 -> 33.54 (Pascal3D+) | Improved |
| add the 3D-aware contrastive loss (ablation) | MPCN · ShapeNet / Pascal3D+ (5-shot) | mean IoU (%): 43.05 -> 47.53 (ShapeNet); 28.76 -> 33.54 (Pascal3D+) | Improved |
| add finetuning on the support set in the second stage (ablation) | MPCN · ShapeNet / Pascal3D+ (5-shot) | mean IoU (%): 45.82 -> 47.53 (ShapeNet); 30.32 -> 33.54 (Pascal3D+) | Improved |
| MPCN vs PADMix with more shots on novel ShapeNet classes | MPCN · ShapeNet novel classes | mean IoU: 0.43 -> 0.44 (1-shot); 0.47 -> 0.51 (10-shot); 0.51 -> 0.54 (25-shot) | Improved |
| MPCN vs Wallace on Pascal3D+ train category | MPCN · Pascal3D+ train (10-shot) | IoU: 0.40 -> 0.37 | Hurt |
| MPCN vs CGCE on ShapeNet cabinet 10-shot | MPCN · ShapeNet cabinet (10-shot) | IoU: 0.71 -> 0.68 | Hurt |
| MPCN vs zero-shot baseline on Pascal3D+ | MPCN · Pascal3D+ (10-shot) | mean IoU: 0.2074 -> 0.3748 | Improved |
Closest papers by meaning
- Few-Shot Generalization for Single-Image 3D Reconstruction via Priors (ICCV 2019 · similarity 0.91)
- Few-Shot Single-View 3-D Object Reconstruction with Compositional Priors (ECCV 2020 · similarity 0.90)
- Single-View 3D Object Reconstruction From Shape Priors in Memory (CVPR 2021 · similarity 0.88)
- Implicit Shape and Appearance Priors for Few-Shot Full Head Reconstruction (TPAMI 2025 · similarity 0.87)
- Learning Shape Priors for Single-View 3D Completion and Reconstruction (ECCV 2018 · similarity 0.87)
- Fostering Generalization in Single-View 3D Reconstruction by Learning a Hierarchy of Local and Global Shape Priors (CVPR 2021 · 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.
