The restworm restores low-quality images that were acquired by sub-optimal imaging settings. It is optimized for images of the nematode worm, C. elegans. The used deep convolutional network is described in the network.py file. The training and test methods including the preparation of data are described below.
The restworm restores images in Input folder by using corresponding images in Ground_truth folder. The folders are designated by the train_input_folder and train_gt_folder in the restworm.py. Paired images must be stored in each folder with the same name.
A trained network is applied for images in Test folder. The filenames are arbitrary. The predicted results are saved in the test_pred_folder.
The restworm relies on the following excellent packages:
- python=3.7.10
- tensorflow-gpu==1.14.0
- keras=2.3.1
- opencv
- tifffile