Research map / 3D vision / Face and shape
Neural Implicit Dictionary Learning via Mixture-of-Expert Training
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 |
|---|---|---|---|
| Increase the number of active experts from k=128 to k=256 (ablation) | NID · CelebA test | PSNR: 28.75 -> 30.26 | Helped |
| Optimize the inferred coding for 5 steps instead of using initialization (ablation) | NID (k=128) · CelebA test | PSNR: 28.75 -> 33.57 | Helped |
| Optimize the inferred coding for 10 steps instead of 5 (ablation) | NID (k=128) · CelebA test | PSNR: 33.57 -> 35.10 | Helped |
| Increase the number of active experts from k=128 to k=256 (ablation) | NID · CT reconstruction, 8 views | PSNR: 16.24 -> 16.77 | Hurt |
| Increase the number of active experts from k=128 to k=256 (ablation) | NID · ShapeNet chair, 10k points | CD: 0.093 -> 0.088 | Helped |
| Increase the number of active experts from k=128 to k=256 (ablation) | NID · ShapeNet chair, 10k points | NC: 0.921 -> 0.945 | Helped |
| Use a learned data-driven function dictionary rather than per-scene SIREN | NID · CelebA test | PSNR: 26.11 -> 37.75 | Helped |
| Use a learned data-driven function dictionary rather than Meta initialization | NID · CT reconstruction, 16 views | PSNR: 21.39 -> 26.32 | Helped |
| Use a learned data-driven function dictionary rather than the strongest listed baseline | NID (k=256) · ShapeNet chair, 10k points | CD: 0.132 -> 0.088 | Helped |
| Use a learned data-driven function dictionary rather than the strongest listed baseline | NID (k=256) · ShapeNet chair, 500k points | CD: 0.051 -> 0.053 | Hurt |
Closest papers by meaning
- Neural Experts: Mixture of Experts for Implicit Neural Representations (NeurIPS 2024 · similarity 0.88)
- Implicit Representations via Operator Learning (ICML 2024 · similarity 0.85)
- Learning Transferable Features for Implicit Neural Representations (NeurIPS 2024 · similarity 0.85)
- SINR: Sparsity Driven Compressed Implicit Neural Representations (CVPR 2025 · similarity 0.85)
- MIRE: Matched Implicit Neural Representations (CVPR 2025 · similarity 0.84)
- Towards Lossless Implicit Neural Representation via Bit Plane Decomposition (CVPR 2025 · 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.