A simple neural net for better image upscaling.
Made by Blaž Rojc (hackguy25) as a seminar assignment for Basics of Artificial Intelligence course.
- Python 3 (3.6.5 was used in creation of the net)
- contents of
requirements.txt
- PyTorch (https://pytorch.org/get-started/locally/)
- optionally:
- CUDA 9.2 (https://developer.nvidia.com/cuda-92-download-archive)
- cuDNN (https://developer.nvidia.com/cudnn), requires registration
- Open
test.py
with a text editor and change the valueskernelSize
,stateName
,src
andout
according to you needs. - Execute the script using
python test.py
in your terminal of choice.
- There are 2 pre-trained net states included,
konv17_1
andkonv25_1
. They requirekernelSize
of 17 and 25, respectively.
- Open
test.py
with a text editor and change the valueskernelSize
,stateName
,src
,processes
,lr
andlrExp
according to you needs. If you are not sure about what to set them to, leavelr
andlrExp
as they are and setprocesses
to number of phisical cores in your processor. - Execute the script using
python train.py
in your terminal of choice.