/SIREN

AI/ML - Project

Primary LanguagePython

SIREN (Implicit Neural Representation with periodic activation functions)

Final project for the Machine Learning and Artificial Intelligence exam at Politecnico di Torino about Implicit Neural Representation with periodic activation functions (SIREN).

Team members

  • Abbamonte Matteo (matteoabbamonte)
  • Gastaldi Paolo (paologastaldi-polito)
  • Gennero Stefano (Stevezbiz)
  • Koudounas Alkis (koudounasalkis)

Contents

  • SIREN implementation
  • Image fitting and Poisson reconstruction
  • Comparison with ReLU
  • Ablation studies
  • SISR (single image super resolution)

Report, slides and code

You can find our report here.

Our presentation slides are here.

All our code is in a Python Notebook format, you can explore it here.

Folders structure

\SIREN
    \docs
        \papers         # papers we referred to  
        \delivery       # our final report, slides and other documents
    \results            # images, graphs and data we obtained, grouped by experiment type
    \src
        \include
            \siren      # content of the original repository we forked from
        \mylibs         # code we exported from the Python Notebook
        \srgan          # pretrained SRGAN values and our config for it
    ...