The code includes IMDeception (core=16+NLA) network and trained parameters.
The network can be found in models folder IMDeception.py file
Which includes
IMDeception Class (proposed network note the core parameter)
GIDB Class (proposed block)
BlockAttention2 (Our efficient implementation of Non-Local Attention Block)
Helper
Gblock and GConv2d classes (Our implementation of grouped conv for faster training, pytorch conv2d with group parameter is too slow to train)

The trained model for IMDeception can be found in
model_zoo folder with imdeception.pth filename

The data folder is cleaned to save bandwidth
The test images must be put in this folder under DIV2K_test_LR folder.

to run the code (once the virtual environment is activated)
"python test_demo.py" must be executed in terminal


The code has been tested in
OS: Ubuntu 20
Torch Version: 1.11.0+cu113
HW: NVIDIA RTX 2080 Super
