Research map / 3D vision / Face and shape
BBoxMaskPose v2: Expanding Mutual Conditioning to 3D
Miroslav Purkrábek, Constantin Kolomiiets, Jiří Matas
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 |
|---|---|---|---|
| add SAM-pose2seg (ablation) | BMPv1 · OCHuman | mask AP: 34.0 -> 40.0 | Improved |
| add SAM-pose2seg (ablation) | BMPv1 · OCHuman | pose AP: 49.2 -> 49.5 | No clear effect |
| add PMPose (ablation) | BMPv1 · OCHuman | pose AP: 49.2 -> 50.2 | Improved |
| add PMPose (ablation) | BMPv1 · OCHuman | mask AP: 34.0 -> 30.8 | Hurt |
| add one extra PMPose pass after mask refinement (BMPv2+) (ablation) | BMPv2 · OCHuman | pose AP: 51.5 -> 55.8 | Improved |
| replace MaskPose with PMPose (ablation) | BMPv1 top-down estimator · OCHuman test | AP: 46.6 -> 48.2 | Improved |
| loop PMPose again (BMPv2+) on COCO (ablation) | BMPv2 · COCO val | AP: 78.8 -> 78.1 | Hurt |
| use keypoint visibility instead of confidence for prompt selection (ablation) | SAM-pose2seg · COCO val | AP: 38.3 -> 44.6 | Improved |
| use keypoint visibility instead of confidence for prompt selection (ablation) | SAM 2.1 · COCO val | AP: 32.7 -> 41.2 | Improved |
| fine-tune SAM for human pose prompts (SAM-pose2seg vs SAM 2.1) (ablation) | SAM · COCO val | AP: 41.2 -> 44.6 | Improved |
| add masks to prompt of SAM-3D-Body (BMPv2) (ablation) | SAM-3D-Body · OCHuman test | 3D pose AP: 46.4 -> 54.8 | Improved |
| replace BMPv1 with BMPv2 for prompting (ablation) | SAM-3D-Body with masks · OCHuman-Pose test | 3D pose AP: 74.9 -> 85.2 | Improved |
Closest papers by meaning
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (NeurIPS 2025 · similarity 0.82)
- The More You See in 2D the More You Perceive in 3D (CVPR 2024 · similarity 0.82)
- Ouroboros3D: Image-to-3D Generation via 3D-aware Recursive Diffusion (CVPR 2025 · similarity 0.82)
- Cue3D: Quantifying the Role of Image Cues in Single-Image 3D Generation (NeurIPS 2025 · similarity 0.82)
- Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention (NeurIPS 2025 · similarity 0.81)
- Neural Signed Distance Function Inference through Splatting 3D Gaussians Pulled on Zero-Level Set (NeurIPS 2024 · similarity 0.81)
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.
