PyNET-QxQ

Setups

Requirements

See requirements.txt.

Data

• Download 3CCD RAW and QxQ dataset and extract it. (These images are stored in 16-bit raw format. A third-party image viewer is required to handle 16-bit RAW images.)

PNG images should be located in ./raw_images/test/. The test folder consists of msc (input images) and gt (target output images). To compute validation metrics, you must exclude QxQ images from test (since QxQ images do not have ground truth images).

Parameters

You should put the parameter file in the ./Experiments/models folder with the name best_acc_model.pth.

Run models

1. Run in cpu-only mode:

python main.py --cpu

2. Run with cuda:

python main.py

Test outputs will be stored in ./Experiments/test_images.

Note

You can update options in option.py to modify settings. For example, to output validation metrics, you can activate --metric option.

python main.py --metric