“Enhancing Low-light Light Field Images with A Deep Compensation Unfolding Network”
You can download the dataset for synthetic LF data from https://drive.google.com/drive/folders/17_4TAdo3AN6qwZLosYCOFUd30evMDDwz?usp=sharing
- Python 3.8.8
- PyTorch 1.13.1
Set the training datapath, and learning rate according to data type. You can also change the batchsize accordingly.
And run 'python train_synf.py'
When training on the L3F dataset, it is advisable to configure the learning rate to 5e-4 for L3F-20, 1e-4 for L3F-50, and 1e-4 for L3F-100, respectively.
For other datasets, we suggest adapting the learning rate selection strategy based on the average brightness level of the dataset. Specifically, a lower learning rate is recommended for datasets with darker overall brightness.
Set the testing datapath.
And run 'python test_synf.py'