Research map / 3D vision / Face and shape
Fixing Implicit Derivatives: Trust-Region Based Learning of Continuous Energy Functions
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 |
|---|---|---|---|
| fuse 72 VOT trackers with learned RBF energy (Our Fusion) instead of best single tracker | Wibergian learning · VOT2018, frames random | IoU: 0.530 -> 0.565 | Helped |
| learned fusion vs best single tracker | Wibergian learning · VOT2018, sequences random | IoU: 0.4643 -> 0.4960 | Helped |
| learned fusion vs median fusion | Wibergian learning · VOT2018, sequences random | IoU: 0.4458 -> 0.4960 | Helped |
| learned fusion vs mean fusion | Wibergian learning · VOT2018, frames random | IoU: 0.238 -> 0.565 | Helped |
| replace Newton step of Samuel et al. with trust-region update | implicit-derivative learning · VOT2018 tracker fusion | IoU | Helped |
| train monocular pose energy parameters end-to-end | Tome et al. monocular model · Human3.6M monocular | average error (mm): 88.4 -> 76.1 | Helped |
| train all parameters vs untrained baseline (monocular) | Ours - Baseline · Human3.6M monocular | average error (mm): 114.2 -> 76.1 | Helped |
| train body model and camera weights only (L2) | multicamera model · Human3.6M multicamera | average error (mm): 91.8 -> 57.8 | Helped |
| also jointly learn grid-searched parameters (L2 loss) (ablation) | multicamera model · Human3.6M multicamera | average error (mm): 57.8 -> 47.7 | Helped |
| use Huber loss instead of L2 with all parameters learned (ablation) | multicamera model · Human3.6M multicamera | average error (mm): 47.7 -> 46.1 | Helped |
| learned parameters vs hand-tuned Tome et al. Huber | multicamera model · Human3.6M multicamera | average error (mm): 52.8 -> 46.1 | Helped |
| monocular learned model vs Martinez et al. | Ours monocular · Human3.6M monocular | average error (mm): 62.9 -> 76.1 | Hurt |
Closest papers by meaning
- Scale-invariant Learning by Physics Inversion (NeurIPS 2022 · similarity 0.83)
- Learning Physics From Video: Unsupervised Physical Parameter Estimation for Continuous Dynamical Systems (CVPR 2025 · similarity 0.81)
- Symmetric Basis Convolutions for Learning Lagrangian Fluid Mechanics (ICLR 2024 · similarity 0.81)
- Diffusion State-Guided Projected Gradient for Inverse Problems (ICLR 2025 · similarity 0.81)
- Bridging Mini-Batch and Asymptotic Analysis in Contrastive Learning: From InfoNCE to Kernel-Based Losses (ICML 2024 · similarity 0.81)
- Physics-Informed Deep Inverse Operator Networks for Solving PDE Inverse Problems (ICLR 2025 · 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.