Research map / 3D vision / Face and shape
Implicit Neural Representation with Multi-Scale Sine Activation
Jufeng Han, Shu Wei, Min Wu, Lina Yu, Weijun Li, Linjun Sun, Hong Qin, Yan Pang
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 |
|---|---|---|---|
| add amplitude modulation (F^-1 scaling) (ablation) | MSA · PDE: Poisson (modified source term) | relative l2 error: 1.31×101 -> 3.49 × 10−5 | Improved |
| add amplitude modulation (F^-1 scaling) (ablation) | MSA · PDE: Poisson (modified source term) | mean-log gradient layer1: 29.01 -> -3.47 | Improved |
| use logarithmic instead of linear frequency spacing (ablation) | MSA-PINN · 2D Helmholtz | relative l2 error | Improved |
| use logarithmic instead of uniform random frequency sampling (ablation) | MSA-PINN · 2D Helmholtz | relative l2 error | Improved |
| use MSA activation instead of SIREN/WIRE/FINER/STAF | MLP INR · 1D multi-scale function fitting | MSE: above 1 × 10−8 -> 7.96×10−14 | Improved |
| use MSA instead of SIREN | MLP INR · BigBuckBunny video (100 frames) | PSNR (dB): 40.19 -> 42.74 | Improved |
| use MSA instead of SIREN | MLP INR · 3D shape SDF: Dragon | IoU: 0.9878 -> 0.9901 | Improved |
| use MSA-PINN instead of PINN | PINN · PDE Heat | relative l2 error: 1.00e+00 -> 1.93e-04 | Improved |
| use MSA-PINN instead of sf-PINN | PINN · PDE Advection | relative l2 error: 9.93e-01 -> 3.35e-04 | Improved |
| use MSA-PINN instead of GP-HM-StM | PINN · PDE Poisson | relative l2 error: 1.26e-04 -> 1.03e-04 | Improved |
| use MSA instead of SIREN (training cost) | MLP INR · BigBuckBunny video | training time (min): 165.39 -> 166.46 | Hurt |
| use MSA instead of ReLU (NTK spectrum) | MLP INR · 1D function fitting | NTK eigenvalue ratio r_remain: ReLU r1 above 60% -> MSA r_remain over 60% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FINER — FINER: Flexible Spectral-bias Tuning in Implicit NEural Representation by Variable-periodic Activation Functions (CVPR 2024) · better on 3 of 3 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
- FINER: Flexible Spectral-bias Tuning in Implicit NEural Representation by Variable-periodic Activation Functions (CVPR 2024 · similarity 0.87)
- Implicit Neural Representations and the Algebra of Complex Wavelets (ICLR 2024 · similarity 0.86)
- MIRE: Matched Implicit Neural Representations (CVPR 2025 · similarity 0.86)
- A sampling theory perspective on activations for implicit neural representations (ICML 2024 · similarity 0.86)
- Tuning the Frequencies: Robust Training for Sinusoidal Neural Networks (CVPR 2025 · similarity 0.86)
- FreSh: Frequency Shifting for Accelerated Neural Representation Learning (ICLR 2025 · 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.
