Home

Awesome

Visitors Python 3.6 Packagist Last Commit Maintenance Contributing Ask Me Anything !

CrossMLP

Cascaded Cross MLP-Mixer GANs for Cross-View Image Translation <br> Bin Ren<sup>1</sup>, Hao Tang<sup>2</sup>, Nicu Sebe<sup>1</sup>. <br> <sup>1</sup>University of Trento, Italy, <sup>2</sup>ETH, Switzerland.<br> In BMVC 2021 Oral. <br> The repository offers the official implementation of our paper in PyTorch.

:t-rex:News! We have updated the proposed CrossMLP(December 9th, 2021)!

Installation

conda create -n crossmlp python=3.6
pip install -r requirement.txt

Dataset Preparation

For Dayton and CVUSA, the datasets must be downloaded beforehand. Please download them on the respective webpages. In addition, we put a few sample images in this code repo data samples. Please cite their papers if you use the data.

Preparing Ablation Dataset. We conduct ablation study in a2g (aerialto-ground) direction on Dayton dataset. To reduce the training time, we randomly select 1/3 samples from the whole 55,000/21,048 samples i.e. around 18,334 samples for training and 7,017 samples for testing. The trianing and testing splits can be downloaded here.

Preparing Dayton Dataset. The dataset can be downloaded here. In particular, you will need to download dayton.zip. Ground Truth semantic maps are not available for this datasets. We adopt RefineNet trained on CityScapes dataset for generating semantic maps and use them as training data in our experiments. Please cite their papers if you use this dataset. Train/Test splits for Dayton dataset can be downloaded from here.

Preparing CVUSA Dataset. The dataset can be downloaded here. After unzipping the dataset, prepare the training and testing data as discussed in our CrossMLP. We also convert semantic maps to the color ones by using this script. Since there is no semantic maps for the aerial images on this dataset, we use black images as aerial semantic maps for placehold purposes.

🌲 Note that for your convenience we also provide download scripts:

bash ./datasets/download_selectiongan_dataset.sh [dataset_name]

[dataset_name] can be:

Training

Run the train_crossMlp.sh, whose content is shown as follows

python train.py --dataroot [path_to_dataset] \
	--name [experiment_name] \
	--model crossmlpgan \
	--which_model_netG unet_256 \
	--which_direction AtoB \
	--dataset_mode aligned \
	--norm batch \
	--gpu_ids 0 \
	--batchSize [BS] \
	--loadSize [LS] \
	--fineSize [FS] \
	--no_flip \
	--display_id 0 \
	--lambda_L1 100 \
	--lambda_L1_seg 1

There are many options you can specify. Please use python train.py --help. The specified options are printed to the console. To specify the number of GPUs to utilize, use export CUDA_VISIBLE_DEVICES=[GPU_ID]. Training will cost about 3 days for dayton , less than 2 days for dayton_ablation, and less than 3 days for cvusa with the default --batchSize on one TITAN Xp GPU (12G). So we suggest you use a larger --batchSize, while performance is not tested using a larger --batchSize

To view training results and loss plots on local computers, set --display_id to a non-zero value and run python -m visdom.server on a new terminal and click the URL http://localhost:8097. On a remote server, replace localhost with your server's name, such as http://server.trento.cs.edu:8097.

Testing

Run the test_crossMlp.sh, whose content is shown as follows:

python test.py --dataroot [path_to_dataset] \
--name crossMlp_dayton_ablation \
--model crossmlpgan \
--which_model_netG unet_256 \
--which_direction AtoB \
--dataset_mode aligned \
--norm batch \
--gpu_ids 0 \
--batchSize 8 \
--loadSize 286 \
--fineSize 256 \
--saveDisk  \ 
--no_flip --eval

By default, it loads the latest checkpoint. It can be changed using --which_epoch.

We also provide image IDs used in our paper here for further qualitative comparsion.

Evaluation

Coming soon

Generating Images Using Pretrained Model

Coming soon

Contributions

If you have any questions/comments/bug reports, feel free to open a github issue or pull a request or e-mail to the author Bin Ren (bin.ren@unitn.it).

Acknowledgments

This source code borrows heavily from SelectionGAN. We also thank the authors X-Fork & X-Seq for providing the evaluation codes. This work was supported by the EU H2020 AI4Media No.951911project and by the PRIN project PREVUE.