Research map / 3D vision / Face and shape
ShapeClipper: Scalable 3D Shape Learning From Single-View Images via Geometric and CLIP-Based Consistency
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 |
|---|---|---|---|
| Add CLIP-based semantic shape consistency (ablation) | ShapeClipper · Pix3D chair test | CD: 0.749 -> 0.618 | Helped |
| Add CLIP-based semantic shape consistency (ablation) | ShapeClipper · Pascal3D+ test | CD: 0.898 -> 0.693 | Helped |
| Add the geometric normal constraint (ablation) | ShapeClipper · Pix3D chair test | CD: 0.766 -> 0.618 | Helped |
| Add the geometric normal constraint (ablation) | ShapeClipper · Pascal3D+ test | CD: 0.952 -> 0.693 | Helped |
| Add outlier dropout for noisy normal losses (ablation) | ShapeClipper · Pix3D chair test | CD: 0.640 -> 0.618 | Helped |
| Add outlier dropout for noisy normal losses (ablation) | ShapeClipper · Pascal3D+ test | CD: 0.707 -> 0.693 | Helped |
| Add outlier dropout for noisy normal losses (ablation) | ShapeClipper · Pascal3D+ test | FS@1: 0.1548 -> 0.1519 | Hurt |
| Use ground-truth viewpoints during training | ShapeClipper · Pix3D chair test | CD: 0.612 -> 0.418 | Helped |
| Increase mask corruption from 0% to 50% | ShapeClipper · Pix3D chair | CD: 0.612 -> 0.763 | Hurt |
| Use CLIP embeddings instead of ImageNet-pretrained ViT embeddings for nearest-neighbor retrieval | CompCars 5-nearest-neighbor model matching · CompCars | model-match percentage: 27.8% -> 51.2% | Helped |
Closest papers by meaning
- Shape-Pose Ambiguity in Learning 3D Reconstruction from Images (AAAI 2021 · similarity 0.89)
- Multiview Aggregation for Learning Category-Specific Shape Reconstruction (NeurIPS 2019 · similarity 0.88)
- Unsupervised Learning of Shape Programs with Repeatable Implicit Parts (NeurIPS 2022 · similarity 0.88)
- Learning to Reconstruct Shapes from Unseen Classes (NeurIPS 2018 · similarity 0.88)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.88)
- 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.