Datasets:
Download croissant.json from RenxingIntelligence/OpenVTON: direct link, hf CLI and curl.
- Browser
- Download file 80.6 kB
-
https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/main/croissant.json
- Command line
-
hf download hf://datasets/RenxingIntelligence/OpenVTON/croissant.json
-
curl -L -o croissant.json https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/main/croissant.json
80.6 kB
{
"@context": {
"@language": "en",
"@vocab": "https://schema.org/",
"citeAs": "cr:citeAs",
"column": "cr:column",
"conformsTo": "dct:conformsTo",
"cr": "http://mlcommons.org/croissant/",
"rai": "http://mlcommons.org/croissant/RAI/",
"data": {
"@id": "cr:data",
"@type": "@json"
},
"dataType": {
"@id": "cr:dataType",
"@type": "@vocab"
},
"dct": "http://purl.org/dc/terms/",
"equivalentProperty": "cr:equivalentProperty",
"examples": {
"@id": "cr:examples",
"@type": "@json"
},
"extract": "cr:extract",
"field": "cr:field",
"fileProperty": "cr:fileProperty",
"fileObject": "cr:fileObject",
"fileSet": "cr:fileSet",
"format": "cr:format",
"includes": "cr:includes",
"isLiveDataset": "cr:isLiveDataset",
"jsonPath": "cr:jsonPath",
"key": "cr:key",
"md5": "cr:md5",
"parentField": "cr:parentField",
"path": "cr:path",
"recordSet": "cr:recordSet",
"references": "cr:references",
"regex": "cr:regex",
"repeated": "cr:repeated",
"replace": "cr:replace",
"samplingRate": "cr:samplingRate",
"sc": "https://schema.org/",
"separator": "cr:separator",
"source": "cr:source",
"subField": "cr:subField",
"transform": "cr:transform",
"prov": "http://www.w3.org/ns/prov#"
},
"@type": "sc:Dataset",
"name": "OpenVTON",
"alternateName": [
"OpenVTON-Bench",
"RenxingIntelligence/OpenVTON"
],
"description": "OpenVTON-Bench contains 99,925 garment-person image pairs with associated masking images and captions: 89,927 training, 4,989 validation, and 5,009 test pairs. Sampling uses 20 visual clusters (C0-C19) from K-Means over DINOv3 garment embeddings. The category column separately contains 14 coarse labels extracted from Gemini captions. This Croissant file describes every published Parquet shard at a fixed data revision; record sets with the same split prefix are concatenated to obtain that split.",
"url": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON",
"license": "https://creativecommons.org/licenses/by-nc/4.0/",
"conformsTo": "http://mlcommons.org/croissant/1.1",
"version": "1.0.0+hf.afe4443ab7961cabe00ec3122809c89111039cde",
"cr:sdVersion": "1.0.0",
"dateCreated": "2026-03-16T06:18:57Z",
"dateModified": "2026-10-09",
"creator": {
"@type": "sc:Organization",
"name": "RenxingIntelligence",
"url": "https://huggingface.co/RenxingIntelligence"
},
"citeAs": "Jin Li et al. OpenVTON-Bench: A Large-Scale High-Resolution Benchmark for Controllable Virtual Try-On Evaluation. arXiv:2601.22725, 2026. https://arxiv.org/abs/2601.22725",
"keywords": [
"virtual try-on",
"benchmark",
"image-to-image",
"DINOv3",
"visual sampling clusters"
],
"rai:dataLimitations": [
"The benchmark evaluates 2D visual try-on quality. It does not establish physical fit, sizing, garment grading, or material physics.",
"Extreme poses, complex multi-layer occlusions, and imaging conditions outside the source distribution may be underrepresented.",
"Caption-derived labels and automated masks may contain errors. Approximate balancing of visual clusters does not imply balanced caption labels or demographic groups."
],
"rai:dataBiases": [
"Source selection from e-commerce and social-media imagery, aesthetic and face-based filtering, and human screening may favor particular clothing styles, presentation conditions, and visible-body configurations.",
"Gemini captions and DINOv3 representations may introduce model-dependent biases. Keyword matching may label an incidental garment mentioned in a caption.",
"Population representativeness and demographic parity have not been established; the release does not contain demographic annotations."
],
"rai:personalSensitiveInformation": [
"The images depict people and may convey apparent personal attributes. Facial regions underwent generative anonymization before release, as described in Section 3.1 of the paper.",
"The released columns are source, mask, target, caption, and category; no participant identity or human-rating records are included in these Parquet files.",
"Users should not use this dataset for identity recovery, biometric identification, face recognition, or surveillance."
],
"rai:dataUseCases": [
"Non-commercial research on controllable virtual try-on, including model evaluation and adaptation using paired garment/person images and masked conditioning inputs.",
"The paper supports model-ranking agreement with human evaluation under a shared five-dimensional visual-quality rubric. This is not a validation of physical fit or unrestricted human-preference prediction.",
"Commercial use is outside CC BY-NC 4.0 authorization. Original asset terms and third-party rights continue to apply."
],
"rai:dataSocialImpact": "The benchmark supports transparent comparisons and analysis of garment fidelity, identity/body consistency, and background preservation. Potential misuse includes deceptive synthetic imagery, unauthorized editing of person images, identity inference, and unrealistic body portrayals. Mitigations described in the paper include facial anonymization, non-commercial licensing, and a takedown mechanism for individuals or rights holders.",
"rai:hasSyntheticData": true,
"rai:dataCollection": "Curated garment-person pairs from the IG Pair dataset associated with IMAGDressing-v1 were combined with publicly available e-commerce imagery. Human screening checked garment-person correspondence and image completeness; the paper describes the source collection and filtering in Section 3 and Appendix A.",
"rai:dataPreprocessingProtocol": [
"Resolution and quality filtering, including aesthetic scoring and face detection, preceded final sampling.",
"DINOv3 garment embeddings were clustered using K-Means with K=20. Final sampling approximately balances these clusters subject to source availability; the released splits contain 89,927/4,989/5,009 pairs.",
"Facial regions were generatively anonymized. GroundingDINO and SAM3 were used for garment localization and masking. The benchmark includes original photographs together with transformed images and generated annotations; it is not a wholly synthetic-photo dataset."
],
"rai:dataAnnotationProtocol": [
"Human verification checked garment-person matches and visible body completeness.",
"Gemini 2.0 Flash generated garment captions using upper-body/lower-body routing and structured prompts. The released category column was derived by deterministic first-match keyword matching, with tshirt mapped to top, knitwear to sweater, and suit to jacket.",
"Qwen-VL-Plus suggested post-hoc garment descriptions for mask prompting; these descriptions did not determine visual-cluster membership. Existing prompts are documented at https://github.com/RenxingIntelligence/OpenVTON-Bench/blob/main/prompts.md."
],
"rai:machineAnnotationTools": [
"Gemini 2.0 Flash",
"DINOv3",
"Qwen-VL-Plus",
"GroundingDINO",
"SAM3"
],
"rai:dataReleaseMaintenancePlan": "The dataset is publicly hosted on Hugging Face; benchmark documentation and implementation are maintained at https://github.com/RenxingIntelligence/OpenVTON-Bench. The paper describes a takedown mechanism for individuals and copyright holders. This metadata is pinned to the data revision named in version and should be regenerated if data shards change.",
"prov:wasDerivedFrom": [
{
"@id": "https://github.com/muzishen/IMAGDressing"
}
],
"prov:wasGeneratedBy": [
{
"@id": "#collection-and-screening",
"@type": "prov:Activity",
"sc:description": "Integration of curated IG Pair garment-person pairs and publicly available e-commerce imagery; quality filtering and human pair verification. See paper Section 3.1-3.2 and Appendix A.2."
},
{
"@id": "#visual-cluster-sampling",
"@type": "prov:Activity",
"sc:description": "DINOv3 garment embedding extraction, K-Means K=20, and cluster-stratified sampling/splitting. Complete counts: https://github.com/RenxingIntelligence/OpenVTON-Bench#data-splits-and-sampling-clusters."
},
{
"@id": "#annotation-and-transformation",
"@type": "prov:Activity",
"sc:description": "Generative face anonymization, Gemini caption generation, deterministic caption-label extraction, and GroundingDINO/SAM3 garment masking. See paper Section 3 and Appendix A; published prompts: https://github.com/RenxingIntelligence/OpenVTON-Bench/blob/main/prompts.md."
}
],
"distribution": [
{
"@type": "cr:FileObject",
"@id": "file-test-00000-of-00002.parquet",
"name": "test-00000-of-00002.parquet",
"description": "Published test split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/test/test-00000-of-00002.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2553937933",
"sha256": "bdd1e5855b6b96afd948a4575397926fdc6efbb410c2757a7d4145375498a557"
},
{
"@type": "cr:FileObject",
"@id": "file-test-00001-of-00002.parquet",
"name": "test-00001-of-00002.parquet",
"description": "Published test split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/test/test-00001-of-00002.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "4191182",
"sha256": "b6b979b5104d6542772240e55cc586403f18b9d847d928249f6c7aa03d7b2911"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00000-of-00018.parquet",
"name": "train-00000-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00000-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2223983159",
"sha256": "557b02f5121d20e48bf48c28c6face30a2c52bce3fb0726a01533ce581e92db2"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00001-of-00018.parquet",
"name": "train-00001-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00001-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2775679372",
"sha256": "eccdf6cf569dabb6536d452376469409069910a5a2c9462484ba5bfee8f84480"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00002-of-00018.parquet",
"name": "train-00002-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00002-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2456863463",
"sha256": "464688c598d77feee7bad1177f6e7c435c4160f3bc786841cdfbcccacb919586"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00003-of-00018.parquet",
"name": "train-00003-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00003-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2587855930",
"sha256": "a00f8b77f1ee3063457415bafa820d06fc4cde9a798b38288f0012bfda6094d1"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00004-of-00018.parquet",
"name": "train-00004-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00004-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2426455080",
"sha256": "2b70d31abc3454e8024adf4bf68cbd9fe994041326a20727042aa0ed70572014"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00005-of-00018.parquet",
"name": "train-00005-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00005-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2451685519",
"sha256": "1bf6168a30b72821bd5ca7d93c9348a777332b243c05bf5a2cbee2b69fe81d54"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00006-of-00018.parquet",
"name": "train-00006-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00006-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2569546648",
"sha256": "36cc7c257fcd0e8cde31b4669a413ef16f7ab972180bfa231408d15ba100fa19"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00007-of-00018.parquet",
"name": "train-00007-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00007-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2352079191",
"sha256": "b0f878e8ef23c75812c32085deb202b0f05c486f5fd7b5ebbfd2a8751b723b9d"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00008-of-00018.parquet",
"name": "train-00008-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00008-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2670909898",
"sha256": "e7632740faac3f6fd806c0bb5b3b1e6642e2fadc152c22685a9ef35f5886d88a"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00009-of-00018.parquet",
"name": "train-00009-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00009-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2802453098",
"sha256": "2ed09a8eb21826b30ebe5e7fa8f9705652780239be6be7f84d336ba0a11fde04"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00010-of-00018.parquet",
"name": "train-00010-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00010-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2692008184",
"sha256": "2a369a9f0568aa9f4d9271e076c302c93fd65dbcd72e36a76225b5374108f986"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00011-of-00018.parquet",
"name": "train-00011-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00011-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2789328885",
"sha256": "91e56a08d866c4b7c2b8258358015ccd7e8dd2ec6554f420dd4cb5b2c2d968da"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00012-of-00018.parquet",
"name": "train-00012-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00012-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2426247269",
"sha256": "b4dd54009794a9b47500572aaf8b5a62b61986e954f32fc4f88036f474a401a2"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00013-of-00018.parquet",
"name": "train-00013-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00013-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2514738691",
"sha256": "97f227891b111d39692122754b56b95fd7a44fd8150c56502d6672201d1ea0e8"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00014-of-00018.parquet",
"name": "train-00014-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00014-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2741484482",
"sha256": "c88624b152e771d7d72698b306e90bdd6545fbd78e702b1b03b52e5980e6fe3a"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00015-of-00018.parquet",
"name": "train-00015-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00015-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2662963705",
"sha256": "9f5d83afb66ebe3eea20e33c547b0b5ca53297c7110aa5b6d512cc305855adaa"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00016-of-00018.parquet",
"name": "train-00016-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00016-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2382833481",
"sha256": "28209af3d3e43ff34cd3192f6c72287b36997c0225ec9563969f28d028cb503c"
},
{
"@type": "cr:FileObject",
"@id": "file-train-00017-of-00018.parquet",
"name": "train-00017-of-00018.parquet",
"description": "Published train split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/train/train-00017-of-00018.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2354622098",
"sha256": "6ca82b0edbff869e6c5e37c29e11375d37e8a807e0da6345cec1e1f74b003d82"
},
{
"@type": "cr:FileObject",
"@id": "file-validation-00000-of-00001.parquet",
"name": "validation-00000-of-00001.parquet",
"description": "Published validation split shard.",
"contentUrl": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON/resolve/afe4443ab7961cabe00ec3122809c89111039cde/validation/validation-00000-of-00001.parquet",
"encodingFormat": "application/x-parquet",
"contentSize": "2527219877",
"sha256": "69f097c36a091dba53f46f845ebf72d6c38bdf02822e70790a21a6f4813e33f4"
}
],
"recordSet": [
{
"@type": "cr:RecordSet",
"@id": "test-00000-of-00002",
"name": "test-00000-of-00002",
"description": "test split, physical shard test/test-00000-of-00002.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "test-00000-of-00002/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-test-00000-of-00002.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "test-00000-of-00002/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-test-00000-of-00002.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "test-00000-of-00002/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-test-00000-of-00002.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "test-00000-of-00002/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-test-00000-of-00002.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "test-00000-of-00002/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-test-00000-of-00002.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "test-00001-of-00002",
"name": "test-00001-of-00002",
"description": "test split, physical shard test/test-00001-of-00002.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "test-00001-of-00002/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-test-00001-of-00002.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "test-00001-of-00002/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-test-00001-of-00002.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "test-00001-of-00002/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-test-00001-of-00002.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "test-00001-of-00002/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-test-00001-of-00002.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "test-00001-of-00002/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-test-00001-of-00002.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00000-of-00018",
"name": "train-00000-of-00018",
"description": "train split, physical shard train/train-00000-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00000-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00000-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00000-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00000-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00000-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00000-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00000-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00000-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00000-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00000-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00001-of-00018",
"name": "train-00001-of-00018",
"description": "train split, physical shard train/train-00001-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00001-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00001-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00001-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00001-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00001-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00001-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00001-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00001-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00001-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00001-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00002-of-00018",
"name": "train-00002-of-00018",
"description": "train split, physical shard train/train-00002-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00002-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00002-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00002-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00002-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00002-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00002-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00002-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00002-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00002-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00002-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00003-of-00018",
"name": "train-00003-of-00018",
"description": "train split, physical shard train/train-00003-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00003-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00003-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00003-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00003-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00003-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00003-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00003-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00003-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00003-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00003-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00004-of-00018",
"name": "train-00004-of-00018",
"description": "train split, physical shard train/train-00004-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00004-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00004-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00004-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00004-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00004-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00004-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00004-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00004-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00004-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00004-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00005-of-00018",
"name": "train-00005-of-00018",
"description": "train split, physical shard train/train-00005-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00005-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00005-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00005-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00005-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00005-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00005-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00005-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00005-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00005-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00005-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00006-of-00018",
"name": "train-00006-of-00018",
"description": "train split, physical shard train/train-00006-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00006-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00006-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00006-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00006-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00006-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00006-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00006-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00006-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00006-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00006-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00007-of-00018",
"name": "train-00007-of-00018",
"description": "train split, physical shard train/train-00007-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00007-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00007-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00007-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00007-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00007-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00007-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00007-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00007-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00007-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00007-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00008-of-00018",
"name": "train-00008-of-00018",
"description": "train split, physical shard train/train-00008-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00008-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00008-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00008-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00008-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00008-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00008-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00008-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00008-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00008-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00008-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00009-of-00018",
"name": "train-00009-of-00018",
"description": "train split, physical shard train/train-00009-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00009-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00009-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00009-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00009-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00009-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00009-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00009-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00009-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00009-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00009-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00010-of-00018",
"name": "train-00010-of-00018",
"description": "train split, physical shard train/train-00010-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00010-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00010-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00010-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00010-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00010-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00010-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00010-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00010-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00010-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00010-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00011-of-00018",
"name": "train-00011-of-00018",
"description": "train split, physical shard train/train-00011-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00011-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00011-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00011-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00011-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00011-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00011-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00011-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00011-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00011-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00011-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00012-of-00018",
"name": "train-00012-of-00018",
"description": "train split, physical shard train/train-00012-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00012-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00012-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00012-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00012-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00012-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00012-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00012-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00012-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00012-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00012-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00013-of-00018",
"name": "train-00013-of-00018",
"description": "train split, physical shard train/train-00013-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00013-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00013-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00013-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00013-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00013-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00013-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00013-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00013-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00013-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00013-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00014-of-00018",
"name": "train-00014-of-00018",
"description": "train split, physical shard train/train-00014-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00014-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00014-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00014-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00014-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00014-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00014-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00014-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00014-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00014-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00014-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00015-of-00018",
"name": "train-00015-of-00018",
"description": "train split, physical shard train/train-00015-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00015-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00015-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00015-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00015-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00015-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00015-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00015-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00015-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00015-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00015-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00016-of-00018",
"name": "train-00016-of-00018",
"description": "train split, physical shard train/train-00016-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00016-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00016-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00016-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00016-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00016-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00016-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00016-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00016-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00016-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00016-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "train-00017-of-00018",
"name": "train-00017-of-00018",
"description": "train split, physical shard train/train-00017-of-00018.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "train-00017-of-00018/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00017-of-00018.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00017-of-00018/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00017-of-00018.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00017-of-00018/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-train-00017-of-00018.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00017-of-00018/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00017-of-00018.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "train-00017-of-00018/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-train-00017-of-00018.parquet"
},
"extract": {
"column": "category"
}
}
}
]
},
{
"@type": "cr:RecordSet",
"@id": "validation-00000-of-00001",
"name": "validation-00000-of-00001",
"description": "validation split, physical shard validation/validation-00000-of-00001.parquet. Concatenate all record sets with the same split prefix to obtain the complete split.",
"field": [
{
"@type": "cr:Field",
"@id": "validation-00000-of-00001/source",
"name": "source",
"description": "Reference garment image.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-validation-00000-of-00001.parquet"
},
"extract": {
"column": "source"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "validation-00000-of-00001/mask",
"name": "mask",
"description": "Person image with the target garment region blacked out, used as the masked-person conditioning input.",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-validation-00000-of-00001.parquet"
},
"extract": {
"column": "mask"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "validation-00000-of-00001/target",
"name": "target",
"description": "Person wearing the reference garment (target image).",
"dataType": "sc:ImageObject",
"source": {
"fileObject": {
"@id": "file-validation-00000-of-00001.parquet"
},
"extract": {
"column": "target"
},
"transform": {
"jsonPath": "bytes"
}
}
},
{
"@type": "cr:Field",
"@id": "validation-00000-of-00001/caption",
"name": "caption",
"description": "Garment description generated with Gemini 2.0 Flash.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-validation-00000-of-00001.parquet"
},
"extract": {
"column": "caption"
}
}
},
{
"@type": "cr:Field",
"@id": "validation-00000-of-00001/category",
"name": "category",
"description": "One of 14 coarse labels obtained by deterministic keyword matching over Gemini captions; not a visual cluster ID.",
"dataType": "sc:Text",
"source": {
"fileObject": {
"@id": "file-validation-00000-of-00001.parquet"
},
"extract": {
"column": "category"
}
}
}
]
}
]
}
{
"@context": {
"@language": "en",
"@vocab": "https://schema.org/",
"citeAs": "cr:citeAs",
"column": "cr:column",
"conformsTo": "dct:conformsTo",
"cr": "http://mlcommons.org/croissant/",
"rai": "http://mlcommons.org/croissant/RAI/",
"data": {
"@id": "cr:data",
"@type": "@json"
},
"dataType": {
"@id": "cr:dataType",
"@type": "@vocab"
},
"dct": "http://purl.org/dc/terms/",
"equivalentProperty": "cr:equivalentProperty",
"examples": {
"@id": "cr:examples",
"@type": "@json"
},
"extract": "cr:extract",
"field": "cr:field",
"fileProperty": "cr:fileProperty",
"fileObject": "cr:fileObject",
"fileSet": "cr:fileSet",
"format": "cr:format",
"includes": "cr:includes",
"isLiveDataset": "cr:isLiveDataset",
"jsonPath": "cr:jsonPath",
"key": "cr:key",
"md5": "cr:md5",
"parentField": "cr:parentField",
"path": "cr:path",
"recordSet": "cr:recordSet",
"references": "cr:references",
"regex": "cr:regex",
"repeated": "cr:repeated",
"replace": "cr:replace",
"samplingRate": "cr:samplingRate",
"sc": "https://schema.org/",
"separator": "cr:separator",
"source": "cr:source",
"subField": "cr:subField",
"transform": "cr:transform",
"prov": "http://www.w3.org/ns/prov#"
},
"@type": "sc:Dataset",
"name": "OpenVTON",
"alternateName": [
"OpenVTON-Bench",
"RenxingIntelligence/OpenVTON"
],
"description": "OpenVTON-Bench contains 99,925 garment-person image pairs with associated masking images and captions: 89,927 training, 4,989 validation, and 5,009 test pairs. Sampling uses 20 visual clusters (C0-C19) from K-Means over DINOv3 garment embeddings. The category column separately contains 14 coarse labels extracted from Gemini captions. This Croissant file describes every published Parquet shard at a fixed data revision; record sets with the same split prefix are concatenated to obtain that split.",
"url": "https://huggingface.co/datasets/RenxingIntelligence/OpenVTON",
"license": "https://creativecommons.org/licenses/by-nc/4.0/",
"conformsTo": "http://mlcommons.org/croissant/1.1",
"version": "1.0.0+hf.afe4443ab7961cabe00ec3122809c89111039cde",
"cr:sdVersion": "1.0.0",
"dateCreated": "2026-03-16T06:18:57Z",
"dateModified": "2026-10-09",
"creator": {
"@type": "sc:Organization",
"name": "RenxingIntelligence",
"url": "https://huggingface.co/RenxingIntelligence"
},
"citeAs": "Jin Li et al. OpenVTON-Bench: A Large-Scale High-Resolution Benchmark for Controllable Virtual Try-On Evaluation. arXiv:2601.22725, 2026. https://arxiv.org/abs/2601.22725",
"keywords": [
"virtual try-on",
"benchmark",
"image-to-image",
"DINOv3",
"visual sampling clusters"
],
"rai:dataLimitations": [
"The benchmark evaluates 2D visual try-on quality. It does not establish physical fit, sizing, garment grading, or material physics.",
"Extreme poses, complex multi-layer occlusions, and imaging conditions outside the source distribution may be underrepresented.",
"Caption-derived labels and automated masks may contain errors. Approximate balancing of visual clusters does not imply balanced caption labels or demographic groups."
],
"rai:dataBiases": [
"Source selection from e-commerce and social-media imagery, aesthetic and face-based filtering, and human screening may favor particular clothing styles, presentation conditions, and visible-body configurations.",
"Gemini captions and DINOv3 representations may introduce model-dependent biases. Keyword matching may label an incidental garment mentioned in a caption.",
"Population representativeness and demographic parity have not been established; the release does not contain demographic annotations."
],
"rai:personalSensitiveInformation": [
"The images depict people and may convey apparent personal attributes. Facial regions underwent generative anonymization before release, as described in Section 3.1 of the paper.",
"The released columns are source, mask, target, caption, and category; no participant identity or human-rating records are included in these Parquet files.",
"Users should not use this dataset for identity recovery, biometric identification, face recognition, or surveillance."
],
"rai:dataUseCases": [
"Non-commercial research on controllable virtual try-on, including model evaluation and adaptation using paired garment/person images and masked conditioning inputs.",
"The paper supports model-ranking agreement with human evaluation under a shared five-dimensional visual-quality rubric. This is not a validation of physical fit or unrestricted human-preference prediction.",
"Commercial use is outside CC BY-NC 4.0 authorization. Original asset terms and third-party rights continue to apply."
],
"rai:dataSocialImpact": "The benchmark supports transparent comparisons and analysis of garment fidelity, identity/body consistency, and background preservation. Potential misuse includes deceptive synthetic imagery, unauthorized editing of person images, identity inference, and unrealistic body portrayals. Mitigations described in the paper include facial anonymization, non-commercial licensing, and a takedown mechanism for individuals or rights holders.",
"rai:hasSyntheticData": true,
"rai:dataCollection": "Curated garment-person pairs from the IG Pair dataset associated with IMAGDressing-v1 were combined with publicly available e-commerce imagery. Human screening checked garment-person correspondence and image completeness; the paper describes the source collection and filtering in Section 3 and Appendix A.",
"rai:dataPreprocessingProtocol": [