AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Rotation-Invariant Spherical Watermarking via Third-Order SO(3) Representation Coupling

Pengzhen Chen, Yanwei Liu, Xiaoyan Gu, Antonios Argyriou, Wu Liu, Weiping Wang

ICML 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
replace third-order bispectrum extraction with second-order power spectrum (ablation)TRIAD · 3D rotation, 32-bit payloadbit accuracy (%): 100.0 -> 61.3Hurt
replace bispectrum with power spectrum at 16-bit payload (ablation)TRIAD · 3D rotation, 16-bit payloadbit accuracy (%): 100.0 -> 92.4Hurt
add rotation-based data augmentation to baseline watermarkersplanar baselines (StegaStamp, TrustMark etc.) · random SO(3) rotationsbit accuracyMixed
use bispectrum-based invariant extraction instead of augmentationTRIAD · random SO(3) rotationsbit accuracyImproved
embed only in low degree V4 (ablation)TRIAD · embedding subspace ablationrobustness vs fidelity: robustness >99%, fidelity compromisedMixed
embed only in high degree V16 (ablation)TRIAD · embedding subspace ablationrobustness vs fidelityMixed
use multi-degree Vembed = V6+V8+V14 (ablation)TRIAD · embedding subspace ablationPSNR and robustness: PSNR 39.22, near-perfect recoveryImproved
increase lmax from 16 to 28 and add higher-degree subspaces (ablation)TRIAD · 3D rotationPSNR: 39.22 -> 37.19Hurt
increase lmax from 16 to 28 (ablation)TRIAD · 3D rotationbit accuracy: 1.000 -> 1.000No clear effect
increase lmax from 4 to 24 (ablation)TRIAD · inference latencylatencyNo clear effect
aggressive rotation augmentation for baselines forces stronger embeddingbaseline watermarkers · panorama watermarkingperceptual qualityHurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy