This repository contains code for our NeurIPS 2021 Workshop on Deep Learning and Inverse Problems paper.
|-- references
| |-- gs
| | |-- non-oversampled
| | | |-- u_cifar_gs.npy
| | | |-- u_emnist_gs.npy
| | | |-- u_fmnist_gs.npy
| | | |-- u_mnist_gs.npy
| | | `-- u_svhn_gs.npy
| | `-- oversampled
| | |-- u_cifar.npy
| | |-- u_emnist.npy
| | |-- u_fmnist.npy
| | |-- u_mnist.npy
| | `-- u_svhn.npy
| |-- hyder
| | |-- non-oversampled
| | | |-- u_cifar.npy
| | | |-- u_emnist.npy
| | | |-- u_fmnist.npy
| | | |-- u_mnist.npy
| | | `-- u_svhn.npy
| | `-- oversampled
| | |-- u_celeba.npy
| | |-- u_cifar.npy
| | |-- u_emnist.npy
| | |-- u_fmnist.npy
| | |-- u_mnist.npy
| | `-- u_svhn.npy
| `-- random
| |-- u_ours_noiseless.npy
| |-- u_ours.npy
| |-- u_random_binary.npy
| `-- u_random.npy
|-- data.py
|-- phase-retrieval-with-reference.ipynb
|-- README.md
|-- unrolled-GS.ipynb
`-- util.py
All experiments were conducted with the following package versions:
- numpy==1.19.5
- torch==1.9.0
- torchvision==0.10.0
- matplotlib==3.4.3
- scikit-image==0.17.2
The reference images for the oversampled case dicussed in Hyder et al. [1] were obtained from the official repository.
[1] Rakib Hyder, Zikui Cai, and M Salman Asif. Solving phase retrieval with a learned reference. In European Conference on Computer Vision, pages 425–441. Springer, 2020.