Research map / 3D vision / Human pose estimation
Every Dog Has Its Day, Probably: A Balanced Synthetic Benchmark and Probabilistic Modeling for 3D Dog Pose Estimation
Joo Young Choi, Wonkwang Lee, Juhyeong Seon, Gunhee Kim
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 |
|---|---|---|---|
| replace deterministic regression with rectified-flow probabilistic model (Mode-of-N, N=20) (ablation) | MMDiT estimator trained on balanced synthetic data · Synthetic benchmark | PCK@0.10 (%): 55.06 -> 64.97 | Improved |
| replace deterministic regression with rectified-flow probabilistic model (Mode-of-N, N=20) (ablation) | MMDiT estimator · Animal3D | PCK@0.10 (%): 69.56 -> 87.70 | Improved |
| replace deterministic regression with rectified-flow probabilistic model (Mode-of-N) (ablation) | MMDiT estimator · StanfordExtra | PCK@0.10 (%): 83.62 -> 85.57 | Improved |
| use oracle Best-of-N selection instead of Mode-of-N | flow-based estimator · Synthetic benchmark | PCK@0.10 (%): 64.97 -> 79.88 | Improved |
| train on balanced synthetic data (ours) instead of existing-benchmark data, progressively (OOOO-D to BBBB-D) (ablation) | deterministic estimator · Synthetic benchmark (60K-image subset) | relative PCK-PA@0.10: monotonic decline when each component replaced (Fig. 6; no numbers in text) | Improved |
| use diffusion-based rendering instead of direct rendering (ablation) | deterministic estimator · Synthetic benchmark (60K-image subset) | relative PCK-PA@0.10: largest single drop (Fig. 6; no numbers in text) | Hurt |
| add probabilistic regression on baseline (non-balanced) data (BBBB-P vs BBBB-D) (ablation) | flow-based estimator · Synthetic benchmark (60K-image subset) | relative PCK-PA@0.10: BBBB-P underperforms BBBB-D (Fig. 6; no numbers in text) | Hurt |
| add probabilistic regression on balanced data (OOOO-P vs OOOO-D) (ablation) | flow-based estimator · Synthetic benchmark (60K-image subset) | relative PCK-PA@0.10: gains concentrated in low visibility, front/back azimuths, large breeds (Fig. 6) | Improved |
| train only on synthetic data vs real-trained baselines | flow-based estimator · StanfordExtra | PCK@0.10 (%): 82.88 (BITE) -> 85.57 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- AniMer+ — AniMer: Animal Pose and Shape Estimation Using Family Aware Transformer (CVPR 2025) · better on 6 of 6 shared results
- BITE — BITE: Beyond Priors for Improved Three-D Dog Pose Estimation (CVPR 2023) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- BITE: Beyond Priors for Improved Three-D Dog Pose Estimation (CVPR 2023 · similarity 0.92)
- Who Left the Dogs Out? 3D Animal Reconstruction with Expectation Maximization in the Loop (ECCV 2020 · similarity 0.91)
- Animal3D: A Comprehensive Dataset of 3D Animal Pose and Shape (ICCV 2023 · similarity 0.90)
- Pose Splatter: A 3D Gaussian Splatting Model for Quantifying Animal Pose and Appearance (NeurIPS 2025 · similarity 0.88)
- BEAST3D: animal behavioral analysis and neural encoding from multi-view video via Gaussian splatting (NeurIPS 2026 · similarity 0.86)
- SAM 3D Animal: Promptable Animal 3D Reconstruction from Images in the Wild (NeurIPS 2026 · 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.
