Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/aimagelab/art2real
Art2Real: Unfolding the Reality of Artworks via Semantically-Aware Image-to-Image Translation. CVPR 2019
https://github.com/aimagelab/art2real
art2real computer-vision cvpr2019 deep-learning image-to-image-translation pytorch
Last synced: 8 days ago
JSON representation
Art2Real: Unfolding the Reality of Artworks via Semantically-Aware Image-to-Image Translation. CVPR 2019
- Host: GitHub
- URL: https://github.com/aimagelab/art2real
- Owner: aimagelab
- Created: 2019-04-02T12:22:09.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2023-03-08T07:22:06.000Z (over 1 year ago)
- Last Synced: 2024-04-19T22:46:49.700Z (7 months ago)
- Topics: art2real, computer-vision, cvpr2019, deep-learning, image-to-image-translation, pytorch
- Language: Python
- Size: 5.5 MB
- Stars: 76
- Watchers: 10
- Forks: 6
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Art2Real
This repository contains the reference code for the paper _[Art2Real: Unfolding the Reality of Artworks via Semantically-Aware Image-to-Image Translation](https://arxiv.org/pdf/1811.10666)_ (CVPR 2019).
Please cite with the following BibTeX:
```
@inproceedings{tomei2019art2real,
title={{Art2Real: Unfolding the Reality of Artworks via Semantically-Aware Image-to-Image Translation}},
author={Tomei, Matteo and Cornia, Marcella and Baraldi, Lorenzo and Cucchiara, Rita},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year={2019}
}
```
## Requirements
This code is built on top of the [Cycle-GAN source code](https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix).
The required Python packages are:
* torch>=0.4.1
* torchvision>=0.2.1
* dominate>=2.3.1
* visdom>=0.1.8.3
* [faiss](https://github.com/facebookresearch/faiss/blob/master/INSTALL.md)
## Pre-trained Models
* **Monet2Photo** [[Checkpoints]](https://ailb-web.ing.unimore.it/publicfiles/drive/CVPR%202019%20-%20Art2Real/checkpoints/monet2photo.zip) [[Dataset]](https://ailb-web.ing.unimore.it/publicfiles/drive/CVPR%202019%20-%20Art2Real/datasets/monet2photo.zip)
* **Landscape2Photo** [[Checkpoints]](https://ailb-web.ing.unimore.it/publicfiles/drive/CVPR%202019%20-%20Art2Real/checkpoints/landscape2photo.zip) [[Dataset]](https://ailb-web.ing.unimore.it/publicfiles/drive/CVPR%202019%20-%20Art2Real/datasets/landscape2photo.zip)
* **Portrait2Photo** [[Checkpoints]](https://ailb-web.ing.unimore.it/publicfiles/drive/CVPR%202019%20-%20Art2Real/checkpoints/portrait2photo.zip) [[Dataset]](https://ailb-web.ing.unimore.it/publicfiles/drive/CVPR%202019%20-%20Art2Real/datasets/portrait2photo.zip)
Download pre-trained models and place them under the checkpoint folder. For example, when downloading the monet2photo checkpoints, place them under the folder `./checkpoints/monet2photo/`.
## Test
Run `python test.py` using the following arguments:
| Argument | Possible values |
|------|------|
| `--dataroot` | Dataset root folder containing the `testA` directory |
| `--name ` | `monet2photo`, `landscape2photo`, `portrait2photo` |
| `--num_test ` | Number of test samples |
For example, to reproduce the results of our model for the first 100 test samples of the landscape2photo setting, use:
```
python test.py --dataroot ./datasets/landscape2photo --name landscape2photo --num_test 100
```
## Training
**Note: for simplicity, the released training code does not include the regular update of semantic masks from the generated images. In this code, original painting masks are kept fixed.**
To run the training code, download the following zip folder containing RGB patches of real landscapes, FAISS indexes and masks from Monet and landscape paintings:
* [[data for patch retrieval]](https://ailb-web.ing.unimore.it/publicfiles/drive/CVPR%202019%20-%20Art2Real/data_for_patch_retrieval.zip)
Place it under the root code folder (*i.e.* `./data_for_patch_retrieval`).
Run `python train.py` using the following arguments:
| Argument | Possible values |
|------|------|
| `--dataroot` | Dataset root folder containing the `trainA` and `trainB` directories |
| `--name ` | Name of the experiment. It decides where to store samples and models |
| `--no_flip ` | Since artistic masks are fixed, we do not random flip images during training |
| `--patch_size_1 ` | Height and width of the first scale patches |
| `--stride_1 ` | Stride of the first scale patches |
| `--patch_size_2 ` | Height and width of the second scale patches |
| `--stride_2 ` | Stride of the second scale patches |
| `--patch_size_3 ` | Height and width of the third scale patches |
| `--stride_3 ` | Stride of the third scale patches |
| `--which_mem_bank ` | `./data_for_patch_retrieval` |
| `--artistic_masks_dir ` | `masks_of_artistic_images_monet`, `masks_of_artistic_images_landscape` |
| `--preload_mem_patches ` | If specified, load all RGB patches in memory |
| `--preload_indexes ` | If specified, load all FAISS indexes in memory |
* Required RAM for both RGB patches and FAISS indexes: ~40 GB.
* Specify only `--patch_size_1 ` and `--stride_1 ` to run the single-scale version.
For example, to train the model on the landscape2photo setting, use:
```
python train.py --dataroot ./datasets/landscape2photo --name landscape2photo --no_dropout --display_id 0 --no_flip --niter_decay 100 --no_flip --patch_size_1 16 --stride_1 6 --patch_size_2 8 --stride_2 5 --patch_size_3 --stride_3 4 --which_mem_bank ./data_for_patch_retrieval --artistic_masks_dir masks_of_artistic_images_landscape --preload_mem_patches --preload_indexes
```
