/LKAT-GAN-for-Infrared-Image-Colorization

This is the code for the paper LKAT-GAN_A_GAN_for_Thermal_Infrared_Image_Colorization_Based_on_Large_Kernel_and_Attention_Unet-Transformer

Primary LanguagePython

LKATGAN

  • PyTorch implementation.

Prerequisites

  • python 3.7
  • torch 1.13.1
  • torchvision 0.14.1

Network

image

LK-Unet

image

Data Preparation

KAIST-MS and IRVI dataset. The resolution of all images in the input network is 256*256.

Colorization results

KAIST-MS dataset

image (a) Thermal infrared iamges. (b) CycleGAN. (c) Pix2pix. (d) SCGAN. (e) TICCGAN. (f) PealGAN. (g) MUGAN. (h) LKAT-GAN. (i)RGB images.

IRVI dataset

image (a) Thermal infrared iamges. (b) I2VGAN. (c) TICCGAN. (d) SCGAN. (e) PealGAN. (f) MUGAN. (g) LKAT-GAN. (h) RGB images.

Comparison methods

CycleGAN. SCGAN. TICCGAN. I2VGAN. PealGAN. MUGAN.

Acknowledgments

This code borrows heavily from TICCGAN.