SASNet: Spatially-Adaptive Sinusoidal Networks for INRs
SIRENs are powerful implicit neural representations, but they are highly sensitive to frequency initialization: low frequencies over-smooth fine detail, while high frequencies inject noise into smooth regions. SASNet addresses this with a fixed frequency embedding paired with jointly learned spatially-adaptive masks that localize each sinusoidal neuron’s influence across the domain. The result is sharper reconstructions, less noise, and faster convergence across image fitting, 3D volumes, and SDF reconstruction — consistently outperforming INR baselines.
Jun 1, 2026
Rethinking Amortized Neural Representations for High-Resolution Terrain Elevation Data
A controlled benchmark of amortized neural representations on high-resolution (1 m/pixel) terrain elevation data, and HUVR+SIREN — a hypernetwork with a smooth, analytically differentiable decoder that attains the best height and derivative fidelity with no extra per-tile storage.
May 29, 2026
ImplicitTerrainV2: Wavelet-Guided Spatially Adaptive Neural Terrain Representation
A follow-up to (CVPR 2024 INRV Workshop), advancing terrain INRs toward a compact, efficient neural terrain data format with wavelet-guided spatial adaptivity, derivative-aware supervision, and post-training compression.
May 21, 2026
Structured Pruning in Implicit Neural Representations
We are actively working on the formating of this paper.
Sep 15, 2024
Presentation of ImplicitTerrain @ CVPR2024 INRV Workshop
We present our CVPR 2024 paper "ImplicitTerrain: a Continuous Surface Model for Terrain Data Analysis" at the 1st Workshop on Implicit Neural Representation for Vision
Jun 18, 2024
ImplicitTerrain: a Continuous Surface Model for Terrain Data Analysis
Please find the project page at .
Jun 18, 2024