Reproduce “Modulated Periodic Activations for Generalizable Local Functional Representations” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Ishit Mehta, Michaël Gharbi, Connelly Barnes, Eli Shechtman, Ravi Ramamoorthi, Manmohan Chandraker

The authors present a representation that uses a sine-activation network, a latent-code modulator, and local tiles for signals. An encoder predicts image codes, while shape codes can be optimized at test time. On 50 unseen Div2K images, it reports 38.03 dB PSNR and 15-second encoding, versus 28.48 dB and 60 minutes for global FFN baselines fitted per image, but training protocols differ.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.