MFG-CRT: Multi-Feature Guided Cross-Refinement Transformer for RGB-Guided Thermal Image Super-Resolution

Overview of MFG-CRT

MFG-CRT: Multi-Feature Guided Cross-Refinement Transformer for RGB-Guided Thermal Image Super-Resolution

PREREQUISITES

Prerequisites for MFG-CRT.

OS

AIR Research Framework is supported on Ubuntu 16.04 LTS or above.

Python

It is recommended that use Python 3.7 or greater, which can be installed either through the Anaconda package manager or the Python website.

Pytorch

Recommended that use Pytorch 1.5.0 or above version. Important: EDVR or some models that have dependency on Deformable Convolution Networks feature only works in Pytorch 1.5.0a0+8f84ded.

Pull container image

At the first, pull docker container image. docker pull nvcr.io/nvidia/pytorch:20.03-py3

Clone

git clone https://github.com/DoGunKIM93/MFG-CRT.git

Install some required packages

pip install fast_slic munch IQA_pytorch pillow fast_pytorch_kmeans

Dataset

×16 Guided Thermal Image Super-Resolution Dataset

datasetPath: 'dataset directory path' (in Param.yaml)

Pre-trained

MFG-CRT Pre-trained

pretrainedPath: 'Pre-trained directory path' (in Param.yaml)

Train

At MFG-CRT folder, type following command:

python main.py

Test

At MFG-CRT folder, type following command:

python main.py -it

Acknowledgements

This code is built on AFA-Net, CRAFT-SR.