Averial-visible-to-infrared-image-translation

AVIID dataset and the code of some representative image-to-image methods for our paper Aerial Visible-to-Infrared Image Translation : Methods, Dataset and Baseline

AVIID dataset

Our proposed dataset consists of three sub-datasets, named AVIID-1, AVIID-2 and AVIID-3. The AVIID-1 contains 993 pairs of paired visible-infrared images with a resolution of 434 $\times$ 434. The AVIID-2 contains 1090 pairs of paired visible-infrared images with a resolution of 434 $\times$ 434. The AVIID-3 contains 1280 pairs of paired visible-infrared images with a resolution of 512 $\times$ 512. The targets in this datset are standard vehicles on the road, including cars, buses, vans, and urban off-road vehicles. The dataset can be downlaoded from https://pan.baidu.com/s/1Tj-8PlsGmRlFv_zlm5Rx6Q, the code is 41uv.

Code of Baseline methods

In our paer, we evaluate ten representative image-to-image methods on our AVIID datset, including Pix2Pix, BicycleGAN, CycleGAN, GCGAN, CUT, DCLGAN, UNIT, MUNIT, DRIT and MSGAN. The details of training and testing of these methods in our paer can be seen here.

Requirements

For Pix2Pix, BicycleGAN, CycleGAN, GCGAN, CUT, DCLGAN, UNIT, and MUNIT:

• Python 3.7 or higher
• Pytorch 1.8.0 or higher, torchvison 0.9.0 or higher
• Tensorboard, TensorboardX, Pyyaml, Pillow, dominate, visdom

For DRIT and MSGAN:

• Python 3.6
• Pytorch 0.4.0, torchvision 0.2.0
• Tensorboard, TensorboardX

Pix2Pix

Training and Testing are followed by https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix.

CycleGAN

Training and Testing are followed by https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix.

BicycleGAN

Training and Testing are followed by https://github.com/junyanz/BicycleGAN.

GCGAN

Training and Testing are followed by https://github.com/hufu6371/GcGAN.

DCLGAN

Training and Testing are followed by https://github.com/JunlinHan/DCLGAN.

CUT

Training and Testing are followed by https://github.com/taesungp/contrastive-unpaired-translation.

UNIT

Training and Testing are followed by https://github.com/mingyuliutw/UNIT.

MUNIT

Training and Testing are followed by https://github.com/NVlabs/MUNIT.

DRIT

Training and Testing are followed by https://github.com/HsinYingLee/DRIT.

MSGAN

Training and Testing are followed by https://github.com/HelenMao/MSGAN.