Research map / Architectures and efficiency / RNNs and normalization
Bottom-Up and Top-Down Reasoning With Hierarchical Rectified Gaussians
Check 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 top-down feedback to QP1 | QP1 keypoint model · COFW visible points | Average keypoint localization error: 5.26 -> 4.67 | Helped |
| Add top-down feedback to QP1 | QP1 keypoint model · COFW all points | Average keypoint localization error: 10.06 -> 7.87 | Helped |
| Add top-down feedback to QP1 | QP1 keypoint model · Pascal VOC 2011 Person | PCK, α=0.10: 66.5 -> 68.8 | Helped |
| Add top-down feedback to QP1 | QP1 keypoint model · Pascal VOC 2011 Person | PCK, α=0.20: 78.9 -> 80.8 | Helped |
| Add top-down feedback to QP1 | QP1 keypoint model · MPII-Val | PCKh-0.5, Full: 81.1 -> 82.4 | Helped |
| Use four coordinate-descent passes rather than one (ablation) | Smaller recurrent network · MPII-Val, smaller network | PCKh(.5), Upper Body: 57.8 -> 61.4 | Helped |
| Use four coordinate-descent passes rather than one (ablation) | Smaller recurrent network · MPII-Val, smaller network | PCKh(.5), Full Body: 59.8 -> 63.1 | Helped |
| Increase coordinate-descent passes beyond four (ablation) | Smaller recurrent network · MPII-Val, smaller network | PCKh(.5), Upper Body: 61.4; 58.7; 60.9 | Mixed |
| Increase coordinate-descent passes beyond four (ablation) | Smaller recurrent network · MPII-Val, smaller network | PCKh(.5), Full Body: 63.1; 61.2; 62.6 | Mixed |
| Use two coordinate-descent passes rather than one (ablation) | Smaller recurrent network · MPII-Val, smaller network | PCKh(.5), Upper Body: 57.8 -> 59.6 | Helped |
| Use two coordinate-descent passes rather than one (ablation) | Smaller recurrent network · MPII-Val, smaller network | PCKh(.5), Full Body: 59.8 -> 62.3 | Helped |
| Use top-down feedback rather than a separate structured-output approach | MPII keypoint prediction compared with TB · MPII | Inference time: 86.5 ms vs TB's 157.2 ms | Helped |
Closest papers by meaning
- Spatial Reasoning with Denoising Models (ICML 2025 · similarity 0.85)
- SpatialLadder: Progressive Training for Spatial Reasoning in Vision-Language Models (ICLR 2026 · similarity 0.85)
- Where Am I and What Will I See: An Auto-Regressive Model for Spatial Localization and View Prediction (ICLR 2025 · similarity 0.85)
- SpatialRGPT: Grounded Spatial Reasoning in Vision-Language Models (NeurIPS 2024 · similarity 0.85)
- Forward Learning with Top-Down Feedback: Empirical and Analytical Characterization (ICLR 2024 · similarity 0.84)
- Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach (AAAI 2026 · similarity 0.84)
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.