This repository stores scripts used to run COMASure and its extensions. The models are studied as part of the requirements for the MSc Data Science and Machine Learning dissertation at UCL.
Disclaimers: All the implementations in this repository are made for research exploration and educational purposes only. They are not ready for industrial usages and the effectiveness is not guaranteed. In addition, permission should be obtained from the author before any distribution or reproduction.
Example: Image 010836 of CelebA
Bicubic: PSNR 23.2, SSIM 0.780
Meta Initialization: PSNR 23.4, SSIM 0.785
Example: Image 0808 of DIV2K
Bicubic: PSNR 21.0, SSIM 0.479
Meta Initialization: PSNR 21.2, SSIM 0.494
CoCoNet: https://github.com/paulbricman/python-fuse-coconet
CoCoNet+MAML: https://github.com/tancik/learnit
Fourier Feature Transformation: https://github.com/tancik/fourier-feature-networks
GAN: https://github.com/eriklindernoren/PyTorch-GAN
MAML: https://github.com/cbfinn/maml
SIREN Layer: https://colab.research.google.com/github/vsitzmann/siren/blob/master/explore_siren.ipynb
Please refer to the actual report.