Research map / 3D vision / Face and shape
NAVI: Category-Agnostic Image Collections with High-Quality 3D Shape and Pose Annotations
Varun Jampani, Kevis-kokitsi Maninis, Andreas Engelhardt, Arjun Karpur, Karen Truong, Kyle Sargent, Stefan Popov, Andre Araujo, Ricardo Martin Brualla, Kaushal Patel, Daniel Vlasic, Vittorio Ferrari and 4 more
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 NAVI GT poses instead of COLMAP poses (ablation) | InstantNGP · NAVI multiview | PSNR: 24.04 -> 27.54 | Improved |
| use NAVI GT poses instead of COLMAP poses (ablation) | InstantNGP · NAVI multiview | SSIM: 0.93 -> 0.94 | Improved |
| use NAVI GT poses instead of COLMAP poses (ablation) | InstantNGP · NAVI multiview | LPIPS: 0.079 -> 0.045 | Improved |
| initialize NeROIC with GT poses instead of COLMAP (ablation) | NeROIC · NAVI in-the-wild S_C | PSNR: 19.77 -> 22.75 | Improved |
| initialize NeRS with GT poses instead of Directions (ablation) | NeRS · NAVI in-the-wild S_C | PSNR: 18.67 -> 17.92 | Hurt |
| initialize SAMURAI with GT poses instead of Directions (ablation) | SAMURAI · NAVI in-the-wild S_C | PSNR: 25.34 -> 25.65 | No clear effect |
| GT pose initialization for NeROIC rotation error (ablation) | NeROIC · NAVI in-the-wild S_C | rotation error (deg): 42.11 -> 33.17 | Improved |
| SAMURAI vs NeRS with Directions init | SAMURAI · NAVI in-the-wild S_C | PSNR: 18.67 -> 25.34 | Improved |
| random initialization (GNeRF) vs pose-informed init | GNeRF · NAVI in-the-wild S_C | PSNR: 25.34 -> 8.30 | Hurt |
| SAMURAI vs NeRS for shape reconstruction | SAMURAI · NAVI in-the-wild shape | IoU: 45.1% -> 63.7% | Improved |
| SAMURAI vs NeRS for shape reconstruction | SAMURAI · NAVI in-the-wild shape | Chamfer distance: 5.86 -> 3.65 | Improved |
| learned matching (SuperGlue) vs SIFT + MNN | SuperPoint + SuperGlue · NAVI correspondence multiview | Precision@0.2: 2.8 -> 26.8 | Improved |
Closest papers by meaning
- Shape-Pose Ambiguity in Learning 3D Reconstruction from Images (AAAI 2021 · similarity 0.87)
- A Real World Dataset for Multi-View 3D Reconstruction (ECCV 2022 · similarity 0.86)
- LEAP: Liberate Sparse-View 3D Modeling from Camera Poses (ICLR 2024 · similarity 0.86)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.86)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.86)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 2024 · similarity 0.85)
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.
