/FRAGAN

A Feature Refinement and Adaptive Generative Adversarial Network for Thermal Infrared Image Colorization

Primary LanguagePython

FRAGAN

A Feature Refinement and Adaptive Generative Adversarial Network for Thermal Infrared Image Colorization

Prerequisites

  • python 3.7
  • torch 1.13.1
  • torchvision 0.14.1
  • dominate
  • visdom

Pretrianed

We provide the pretrained model: Google

Trian

python train.py --dataroot [dataset root] --name [experiment_name] --phase train --which_epoch latest

Test

python test.py --dataroot [dataset root] --name [experiment_name] --phase test --which_epoch latest

Colorization results

KAIST dataset

KAIST

FLIR dataset

FLIR

Note

We only uploaded the paired strategy, the remaining strategies can be found in CycleGAN and CUT.

Acknowledgments

This code heavily borrowes from MUGAN.