Research map / 3D vision / Face and shape
COSMO-INR: Complex Sinusoidal Modulation for Implicit Neural Representations
Pandula Thennakoon, Avishka Ranasinghe, Mario De Silva, Buwaneka Epakanda, Roshan Godaliyadda, Mervyn Parakrama Bandara Ekanayake, Vijitha R. Herath
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 complex sinusoidal modulation to the raised cosine activation (ablation) | Raised cosine INR activation · INR image representation | PSNR | Improved |
| replace INCODE with COSMO-RC activation | INR MLP · Kodak image representation (24 images) | PSNR (dB): 35.57 -> 41.24 | Improved |
| replace INCODE with COSMO-RC activation | INR MLP · NeRF Lego test set | PSNR (dB): 26.05 -> 29.50 | Improved |
| replace INCODE with COSMO-RC activation | INR MLP · DIV2K denoising | PSNR (dB): +0.46 dB | Improved |
| replace baselines with COSMO-RC activation | INR MLP · Celtic knots inpainting and Lucy occupancy volume | PSNR / IoU | Improved |
| replace baselines with COSMO-RC activation | INR MLP · DIV2K super-resolution 2x, 4x, 6x | PSNR (dB): FINER 32.94 -> 34.03 at 2x; INCODE 29.96 -> 30.42 at 4x; SIREN 27.31 -> 27.66 at 6x | Improved |
| increase network width from 64 to 512 at depth 4 (ablation) | COSMO-RC · Kodak image 22 representation | PSNR (dB): 29.47 -> 52.00 | Improved |
| increase depth from 2 to 4 hidden layers at width 256 (ablation) | COSMO-RC · Kodak image 22 representation | PSNR (dB): 38.09 -> 39.57 -> 37.19 | Mixed |
| add prior embedder with sigmoid regularizer for activation parameters | COSMO-RC · all image tasks | convergence speed | Improved |
| use COSMO-RC instead of SIREN, FINER, INCODE or WIRE | INR MLP · Kodak image 20 compute cost | train time (s/it): 0.435 -> 3.500 vs INCODE | Hurt |
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 4 of 4 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
- VI^3NR: Variance Informed Initialization for Implicit Neural Representations (CVPR 2025 · similarity 0.91)
- Implicit Representations via Operator Learning (ICML 2024 · similarity 0.90)
- Grids Often Outperform Implicit Neural Representation at Compressing Dense Signals (NeurIPS 2025 · similarity 0.90)
- MIRE: Matched Implicit Neural Representations (CVPR 2025 · similarity 0.89)
- Implicit Neural Representations and the Algebra of Complex Wavelets (ICLR 2024 · similarity 0.89)
- A sampling theory perspective on activations for implicit neural representations (ICML 2024 · similarity 0.89)
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.
