vsitzmann/siren
A neural network architecture using periodic activation functions to represent signals as implicit functions.
Not currently ranked — collecting fresh signals.
star history
This is the official implementation of SIREN (Sinusoidal Representation Networks), a neural network approach that uses periodic activation functions instead of conventional non-linearities. The method enables neural networks to represent complex signals like images and audio directly as functions, and can solve partial differential equations like Poisson’s equation. The paper demonstrates that SIRENs can naturally represent audio, images, and video as well as their derivatives.
Frequently asked
- What is vsitzmann/siren?
- A neural network architecture using periodic activation functions to represent signals as implicit functions.
- Is siren open source?
- Yes — vsitzmann/siren is open source, released under the MIT license.
- What language is siren written in?
- vsitzmann/siren is primarily written in Python.
- How popular is siren?
- vsitzmann/siren has 2k stars on GitHub.
- Where can I find siren?
- vsitzmann/siren is on GitHub at https://github.com/vsitzmann/siren.