zhs1 / vdnet

Variational Denoising Network: Toward Blind Noise Modeling and Removal (NeurIPS, 2019) (Pytorch)

VDNet

Variational Denoising Network: Toward Blind Noise Modeling and Removal (NeurIPS, 2019) arXiv

Requirements and Dependencies

• Ubuntu 16.04, cuda 10.0
• Python 3.6, Pytorch 1.1.0
• More detail (See environment.yml)

Testing

1. For the Non-IID Gaussian Denosing as described in the paper, please run demo_test_simulation.py.
2. For real-wormd image denoising task, please run demo_test_benckmark.py. The model was trained on the SIDD Medium Dataset (320 noisy and clean paris).

Training

1. Prepare the dataset following the code in the floder datasets. Data link: Waterloo Exploration Database, CBSD432 and CImageNet400.
2. Train VDN using train_simulation.py or train_benchmark.py.