{"id":13487773,"url":"https://github.com/cuiaiyu/street-tryon-benchmark","last_synced_at":"2025-03-27T23:31:40.610Z","repository":{"id":209255992,"uuid":"723595185","full_name":"cuiaiyu/street-tryon-benchmark","owner":"cuiaiyu","description":"[WACV'25] StreetTryOn: A Benchmark for In-the-Wild Virtual Try-On and Cross-Domain Virtual Try-On","archived":false,"fork":false,"pushed_at":"2024-11-19T03:23:48.000Z","size":10194,"stargazers_count":125,"open_issues_count":3,"forks_count":10,"subscribers_count":9,"default_branch":"main","last_synced_at":"2024-11-19T04:25:01.545Z","etag":null,"topics":["image-editing","image-generation","virtual-try-on"],"latest_commit_sha":null,"homepage":"https://arxiv.org/pdf/2311.16094.pdf","language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/cuiaiyu.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2023-11-26T06:47:52.000Z","updated_at":"2024-11-19T03:23:51.000Z","dependencies_parsed_at":"2024-10-30T23:31:02.223Z","dependency_job_id":"2eefd31f-d813-4995-bc7c-a3da9cae42e2","html_url":"https://github.com/cuiaiyu/street-tryon-benchmark","commit_stats":null,"previous_names":["cuiaiyu/street-tryon-benchmark"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cuiaiyu%2Fstreet-tryon-benchmark","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cuiaiyu%2Fstreet-tryon-benchmark/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cuiaiyu%2Fstreet-tryon-benchmark/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/cuiaiyu%2Fstreet-tryon-benchmark/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/cuiaiyu","download_url":"https://codeload.github.com/cuiaiyu/street-tryon-benchmark/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":245944020,"owners_count":20697945,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["image-editing","image-generation","virtual-try-on"],"created_at":"2024-07-31T18:01:03.647Z","updated_at":"2025-03-27T23:31:40.600Z","avatar_url":"https://github.com/cuiaiyu.png","language":"Jupyter Notebook","funding_links":[],"categories":["Personalized Restoration","Datasets for Virtual Try-on"],"sub_categories":[],"readme":"# Street TryOn Dataset/Benchmark\n[\\[Paper\\]](https://arxiv.org/pdf/2311.16094.pdf) \n[\\[Project Page\\]](https://cuiaiyu.github.io/StreetTryOn/)\n\n(WACV'25)\n\n\n__StreetTryOn__, the new in-the-wild Virtual Try-On dataset, consists of 12, 364 and 2, 089 street\nperson images for training and validation, respectively. It is derived from the large fashion retrieval\ndataset [DeepFashion2](https://github.com/switchablenorms/DeepFashion2), from which we\nfilter out over 90% of DeepFashion2 images that are infeasible for try-on tasks (e.g., non-frontal view, large occlusion,\ndark environment, etc.).\nCombining with the garment and person images in [VITON-HD](https://github.com/shadow2496/VITON-HD), we obtain a comprehensive suite of in-domain\nand cross-domain try-on tasks that have garment and person inputs from various sources, including Shop2Model,\nModel2Model, Shop2Street, and Street2Street.\n\nThis is the official release of the __StreetTryOn__ Dataset and more details can be found in our paper [Street TryOn: Learning In-the-wild Virtual Try-On from Unpaired Person Images](https://arxiv.org/pdf/2311.16094.pdf).\n\n    \n![image](cover_images/teaser_right.png)\n\nFive Virtual Try-On tasks are covered:\n- In-domain Tasks: __Shop2Model Test, Model2Model Test, Street2Street Test__ (New)\n- Cross-domain Tasks: __Shop2Street Test__ (New), __Model2Street Test__ (New)\n\n## Street TryOn Dataset\nStreet TryOn Dataset contains __unpaired__ __in-the-wild person images__ that can be used for virtual try-on tasks. Street TryOn Dataset consists of 12,364 and 2089 images filtered from [Deepfashion2 Dataset](https://github.com/switchablenorms/DeepFashion2) for training and validation.\n\n\nWe release all the annotations mentioned in [our paper](https://arxiv.org/pdf/2311.16094.pdf). Note for images: we provide scripts to extract them from DeepFashion2 dataset. Please follow the below steps to download the dataset into your datapath `$DATA`. \n\n### Licenses\nSince this dataset is derived from  [Deepfashion2 Dataset](https://github.com/switchablenorms/DeepFashion2), the same license is inherited.\n\n### Downloading Steps\n1. Obtain the data access of DeepFashion2 from [its official release](https://github.com/switchablenorms/DeepFashion2#download-the-data) for a __zip password__ which will be used later to unzip the images.\n2. Clone this repo under `$DATA` by \n```sh\ngit clone https://github.com/cuiaiyu/street-tryon-benchmark\n ```\n2. Download the released Street TryOn Data annotations from [this link](https://drive.google.com/drive/folders/1IxcCiG4FID1uRoMdm2wSapfNsYBCPXDH?usp=sharing) and unzip it under `$DATA` as `$DATA/street_tryon`\n3. Download, filter and process the images from DeepFasshion2 by running the below script \n```sh\n# The password obtained in Step 1 will be used here.\nsh street-tryon-benchmark/get_street_images.sh\n```\n4. Move the tutorial notebook under `$DATA`\n```\nmv street-tryon-benchmark/street_tryon_tutorial.ipynb .\n```\n\n\n\n4. After that, you should have the data as:\n```\n- $DATA\n    - street_tryon/\n        - train/\n            - image/\n            - densepose/\n            - raw_bbox/\n            - ...\n        - validation/\n            - image/\n            - densepose/\n            - raw_bbox/\n            - ...\n        - annotations/\n            - street2street_test_pairs_top.csv\n            - street2street_test_pairs_dress.csv\n            - ...\n    - street-tryon-benchmark/\n        - ...\n    - street_tryon_tutorial.ipynb\n        \n```\n\n## Set up VITON-HD Dataset for Cross-domain Virtual Try-On Test\nTo run the cross-domain virtual try-on test, please also download the [VITON-HD dataset](https://github.com/shadow2496/VITON-HD#dataset) from its official release and unzip it under `$DATA`. The `$DATA` directory should look like\n```\n- $DATA\n    - street_tryon/\n        - train/\n            - ...\n        - validation/\n            - ...\n        - annotations/\n            - ...\n    - zalando/ (VITON-HD)\n        - train/\n            - ...\n        - test/\n            - ...\n    - street-tryon-benchmark/\n        - ...\n    - street_tryon_tutorial.ipynb\n```\n### Additional annotations for VITON-HD\n\nWe also release the full DensePose annotations for both human and garment images in VITON-HD:\n- __Human DensePose__ ([Download link](https://drive.google.com/drive/folders/1Ha_Xzl9QZ22hx_1kQ-DymPrNYYtWhg0U?usp=sharing)): obtained by the official model of [detectron2-DensePose](https://github.com/facebookresearch/detectron2/blob/main/projects/DensePose/doc/GETTING_STARTED.md)\n- __Garment DensePose__ ([Download link](https://drive.google.com/drive/folders/1Ha_Xzl9QZ22hx_1kQ-DymPrNYYtWhg0U?usp=sharing)): obtained by the reimplemented model of [this method by Cui et al. 2023](https://arxiv.org/abs/2303.17688)\n\nPlease download the data and put them in `zalando/train` and `zalando/test` respectively.\n\n*If you find the additional annotation is helpful, please consider citing the original detection methods.*\n\n\n## Load Data for Multiple Tests\nWe provide a PyTorch dataloader to load data from the same domain or cross domain flexibly. \n\nAfter `VITON-HD` and `street-tryon` datasets are set up, one can run the following code to load the data for the proposed tests in the paper:\n\n\n```python\nfrom street_tryon_benchmark.dataset import GeneralTryOnDataset\n\n\ndef get_dataset_by_task(task):\n    if task == 'shop2model':\n        config_path = \"street_tryon_benchmark/configs/shop2model.yaml\"\n    elif task == 'shop2street':\n        config_path = \"street_tryon_benchmark/configs/shop2street.yaml\"\n    elif task == 'model2model':\n        config_path = \"street_tryon_benchmark/configs/model2model.yaml\"\n    elif task == 'model2street':\n        config_path = \"street_tryon_benchmark/configs/model2street.yaml\"\n    elif task == 'street2street-top':\n        config_path = \"street_tryon_benchmark/configs/street2street_top.yaml\"\n    elif task == 'street2street-dress':\n        config_path = \"street_tryon_benchmark/configs/street2street_dress.yaml\"\n    else:\n        raise NotImplementedError\n\n\n    with open(config_path, \"r\") as f:\n        data_config = yaml.safe_load(f)\n\n    return GeneralTryOnDataset(\".\", config=data_config, split='test')\n\n# create dataset for street2street task\ndataset = get_dataset_by_task('street2street-top')\n\n# check data\ncurr = dataset[0]\n\n# get person-related data\npimg, piuv, pseg = curr['pimg'], curr['piuv'], curr['pseg']\n\n# get garment-related data\ngimg, giuv, gseg = curr['gimg'], curr['giuv'], curr['gseg']\n```\n\nWe also provide a [notebook](street_tryon_tutorial.ipynb) to play with this dataloader.\n\nNote:\n- The default dataset configuration is loading ATR segmentation for street images and loading the provided segmentation for VITON-HD. If you need either in different format, you can change the `segm_dir` or `garment_segm_dir` in the corresponding ``.yaml`` config file with your new datapath.\n\n## Licenses\nWe inherit the licenses from both [DeepFashion2 Dataset](https://github.com/switchablenorms/DeepFashion2) and [VITON-HD](https://github.com/shadow2496/VITON-HD#license). The usage of the data and code has to meet the requirements of both licenses.\n\nNo commerical usage is allowed.\n\n## Citations\nIf you find this work helpful, please cite us as:\n```\n@article{cui2023street-tryon,\n  title={Street TryOn: Learning In-the-Wild Virtual Try-On from Unpaired Person Images},\n  author={Cui, Aiyu and Mahajan, Jay and Shah, Viraj and Gomathinayagam, Preeti and Lazebnik, Svetlana},\n  journal={arXiv preprint arXiv:2311.16094},\n  year={2023}\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fcuiaiyu%2Fstreet-tryon-benchmark","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fcuiaiyu%2Fstreet-tryon-benchmark","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fcuiaiyu%2Fstreet-tryon-benchmark/lists"}