Croissant: Haoran Qin affiliated with Simon Fraser University and Westlake University
41c1f7d verified | { | |
| "@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" | |
| }, | |
| "@type": "sc:Dataset", | |
| "name": "RoboCloth", | |
| "description": "RoboCloth is a dataset of 500 real cloth materials captured under flexible, calibrated camera and lighting configurations, where both camera and light are robotically controlled during acquisition. Each material is imaged in approximately 580 high-dynamic-range frames spanning a wide view-light hemisphere. Camera poses are refined via structure-from-motion and aligned to the robot frame, yielding per-material sparse surface geometry and structured per-point reflectance observations ready for neural material training.\n\nThe release contains, for every material, the sample-masked linear HDR frames, the structured per-point observation tensor, the sparse surface point cloud and the full per-frame rig log, plus dataset-level radiometric and emitter calibration and nested train/test splits. It is intended for BRDF / SVBRDF and BTF estimation, neural appearance modelling, multi-view multi-light inverse rendering and material retrieval research.", | |
| "url": "https://theialab.github.io/robocloth/", | |
| "license": "https://creativecommons.org/licenses/by/4.0/", | |
| "conformsTo": "http://mlcommons.org/croissant/1.0", | |
| "version": "2.0.0", | |
| "datePublished": "2026-09-02", | |
| "isLiveDataset": false, | |
| "citeAs": "@misc{li2026robocloth,\n title = {RoboCloth: A Large-Scale Real Cloth Material Dataset for Neural Reflectance Reconstruction},\n author = {Li, Zhen and Qin, Haoran and Di Sario, Francesco and Rebain, Daniel and Tagliasacchi, Andrea},\n year = {2026}\n}", | |
| "creator": [ | |
| { | |
| "@type": "Person", | |
| "name": "Zhen Li", | |
| "affiliation": { | |
| "@type": "Organization", | |
| "name": "Simon Fraser University" | |
| } | |
| }, | |
| { | |
| "@type": "Person", | |
| "name": "Haoran Qin", | |
| "affiliation": [ | |
| { | |
| "@type": "Organization", | |
| "name": "Simon Fraser University" | |
| }, | |
| { | |
| "@type": "Organization", | |
| "name": "Westlake University" | |
| } | |
| ] | |
| }, | |
| { | |
| "@type": "Person", | |
| "name": "Francesco Di Sario", | |
| "affiliation": [ | |
| { | |
| "@type": "Organization", | |
| "name": "University of Torino" | |
| }, | |
| { | |
| "@type": "Organization", | |
| "name": "Simon Fraser University" | |
| } | |
| ] | |
| }, | |
| { | |
| "@type": "Person", | |
| "name": "Daniel Rebain", | |
| "affiliation": [ | |
| { | |
| "@type": "Organization", | |
| "name": "University of British Columbia" | |
| }, | |
| { | |
| "@type": "Organization", | |
| "name": "Wayve Labs" | |
| } | |
| ] | |
| }, | |
| { | |
| "@type": "Person", | |
| "name": "Andrea Tagliasacchi", | |
| "affiliation": [ | |
| { | |
| "@type": "Organization", | |
| "name": "Simon Fraser University" | |
| }, | |
| { | |
| "@type": "Organization", | |
| "name": "Wayve Labs" | |
| }, | |
| { | |
| "@type": "Organization", | |
| "name": "University of Toronto" | |
| } | |
| ] | |
| } | |
| ], | |
| "keywords": [ | |
| "BRDF", | |
| "BTF", | |
| "cloth", | |
| "materials", | |
| "reflectance", | |
| "material capture", | |
| "multi-view", | |
| "multi-light", | |
| "HDR", | |
| "robotics", | |
| "inverse rendering", | |
| "computer graphics" | |
| ], | |
| "rai:dataCollection": "Captures were performed with a robotic rig: two arms carry a machine-vision camera and a single LED emitter over a turntable holding a flat-mounted cloth sample, so the view-light hemisphere is sampled densely rather than on a fixed gantry grid. Each material yields approximately 580 frames, captured as 14-bit Bayer mosaics and released as full-frame (3072x2048) 16-bit linear RGB PNGs masked to the sample polygon (Menon-2007 demosaicing, white-balance gains applied, no gamma). Camera poses come from a COLMAP structure-from-motion reconstruction registered into the turntable-aligned robot frame by a Umeyama alignment against the robot-logged camera poses; frames failing a 16 mm alignment gate are dropped. The same reconstruction yields the sparse surface point cloud, cropped to the physical footprint of the sample. Per-point radiance observations across all registered frames are then assembled into one structured tensor per material. Camera intrinsics are stored per material in hdr_crop_bboxes.json; the LED angular falloff and the per-material exposure-regime index are released under globals/; hand-eye, turntable-axis and radiometric-scale constants are in the code repository (configs/renderer/rig_constants.yaml).", | |
| "rai:dataAnnotationProtocol": "No human annotation. All metadata (camera poses, light positions, per-pixel observations, point cloud) is derived from the calibration pipeline.", | |
| "rai:dataLimitations": "- Single capture rig: its lens model, LED angular falloff and geometric layout are baked into the data.\n- Samples are flat-mounted; cloth deformation and drape are not part of the capture.\n- Very specular or very dark materials can still clip at grazing angles despite the 16-bit linear encoding.\n- Geometry is a sparse structure-from-motion point cloud, not a dense scan.\n- Frames that structure-from-motion never registered are listed per material in unmatched_scan_ids.json and are absent from the observation tensor.", | |
| "rai:dataBiases": "- The material collection reflects the fabrics available to the capture lab and is not a representative cross-section of global textile diversity.\n- Reflected hemisphere only: no transmissive setups and no dedicated subsurface-scattering captures.\n- HDR capture, while wide-range, may saturate on very specular or very dark materials.", | |
| "rai:personalSensitiveInformation": "None. Data consists exclusively of cloth material captures. No people, no faces, no identifiable subjects, no personally-identifying metadata.", | |
| "rai:dataUseCases": "- BRDF / SVBRDF and BTF estimation\n- Neural appearance models usable as online or offline shaders\n- Multi-view multi-light inverse rendering\n- Photometric stereo and surface-reflectance benchmarks\n- Material classification or retrieval research", | |
| "rai:dataSocialImpact": "- Intended for graphics, vision, and inverse-rendering research.\n- No known dual-use risk: cloth material captures are physical-object\n measurements with no human subjects, no personally-identifying\n metadata, and no operational security implications.\n- Indirect downstream uses might include realistic cloth rendering for\n games, films, or virtual try-on; the dataset itself does not enable\n surveillance or harm.", | |
| "rai:hasSyntheticData": false, | |
| "distribution": [ | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-hdr-tar", | |
| "name": "materials-hdr-tar", | |
| "description": "Tar archive (hdr/*.png) of the registered multi-view multi-light frames as full-size 3072x2048 16-bit linear PNGs, masked to the sample polygon.", | |
| "encodingFormat": "application/x-tar", | |
| "includes": "materials/*/hdr.tar" | |
| }, | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-observations-structured-npz", | |
| "name": "materials-observations-structured-npz", | |
| "description": "NumPy npz containing structured per-(point,camera,light) observations: xyz (m), point_ids, rgbs (K,V,3) uint16, cam_pos and light_pos (robot-logged gripper and LED positions, mm).", | |
| "encodingFormat": "application/zip", | |
| "includes": "materials/*/observations_structured.npz" | |
| }, | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-point-positions-npz", | |
| "name": "materials-point-positions-npz", | |
| "description": "NumPy npz with sparse 3D point cloud: point_ids, positions.", | |
| "encodingFormat": "application/zip", | |
| "includes": "materials/*/point_positions.npz" | |
| }, | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-rotated-camera-json", | |
| "name": "materials-rotated-camera-json", | |
| "description": "Per-camera pose array: overall_id, camera_id, position[3] (mm), rotation_matrix[3x3] (SfM-refined, turntable-aligned frame; includes the Umeyama scale).", | |
| "encodingFormat": "application/json", | |
| "includes": "materials/*/rotated_camera.json" | |
| }, | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-scan-log-json", | |
| "name": "materials-scan-log-json", | |
| "description": "Per-scan log: id, scan_id, camera_id, light_id, position[3], rotation_matrix[3x3], position_light[3], rotation_matrix_light[3x3], servo_angle, servo_angle_light, turn_angle, filename (mm, deg).", | |
| "encodingFormat": "application/json", | |
| "includes": "materials/*/scan_log.json" | |
| }, | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-point-metadata-json", | |
| "name": "materials-point-metadata-json", | |
| "description": "Per-material counts: num_points, num_observations (observations_folder present in some materials).", | |
| "encodingFormat": "application/json", | |
| "includes": "materials/*/point_metadata.json" | |
| }, | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-bbox-json", | |
| "name": "materials-bbox-json", | |
| "description": "Sample bounding box: bbox_min, bbox_max, bbox_center, bbox_size, num_points.", | |
| "encodingFormat": "application/json", | |
| "includes": "materials/*/bbox.json" | |
| }, | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-hdr-crop-bboxes-json", | |
| "name": "materials-hdr-crop-bboxes-json", | |
| "description": "Camera intrinsics plus the per-frame pixel box and polygon of the sample mask applied to the released frames (dilate_px).", | |
| "encodingFormat": "application/json", | |
| "includes": "materials/*/hdr_crop_bboxes.json" | |
| }, | |
| { | |
| "@type": "cr:FileSet", | |
| "@id": "materials-unmatched-scan-ids-json", | |
| "name": "materials-unmatched-scan-ids-json", | |
| "description": "Scan ids that lacked a camera/light correspondence (small or empty for clean materials).", | |
| "encodingFormat": "application/json", | |
| "includes": "materials/*/unmatched_scan_ids.json" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-camera-factor-json", | |
| "name": "globals-camera-factor-json", | |
| "description": "Per-material exposure-regime index (camera_factor_segments: id ranges -> factor 1, 2 or 3).", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/camera_factor.json", | |
| "encodingFormat": "application/json", | |
| "contentSize": "3923 B", | |
| "sha256": "442a5494166ae9b61f428920f6b4ec4178e5aa671f521a674c37fa8a12a5451a" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-emitter-calibration-json", | |
| "name": "globals-emitter-calibration-json", | |
| "description": "Light emitter intensity profile by angular position (degrees).", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/emitter_calibration.json", | |
| "encodingFormat": "application/json", | |
| "contentSize": "3027 B", | |
| "sha256": "78e263ce2c7095c1abd9cf4ab194407141974912a56c4199fe40ca71de1a8d41" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-sample-size-json", | |
| "name": "globals-sample-size-json", | |
| "description": "Physical sample width and length (m) per material-id range, used to crop the point cloud.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/sample_size.json", | |
| "encodingFormat": "application/json", | |
| "contentSize": "7142 B", | |
| "sha256": "190f83e9a86da3ead2f435017d0506898ae001d5b9fa66fee2d24711ef54a81f" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-training-list-100-txt", | |
| "name": "globals-training-list-100-txt", | |
| "description": "Training material ids for the 100-material subset.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/training_list_100.txt", | |
| "encodingFormat": "text/plain", | |
| "contentSize": "299 B", | |
| "sha256": "7cf26c4900a3760ad1489a2e890f81aaa9da0dad87fec946c220fd1bce9c4fa7" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-training-list-300-txt", | |
| "name": "globals-training-list-300-txt", | |
| "description": "Training material ids for the 300-material subset.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/training_list_300.txt", | |
| "encodingFormat": "text/plain", | |
| "contentSize": "901 B", | |
| "sha256": "74da0f01bdfde6a57180f51799bf85666117e6863145dc27048bdf720954c552" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-training-list-442-txt", | |
| "name": "globals-training-list-442-txt", | |
| "description": "Training material ids for the 442-material subset.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/training_list_442.txt", | |
| "encodingFormat": "text/plain", | |
| "contentSize": "1330 B", | |
| "sha256": "4c8399b8daddefe8e3b4c28e8105608e4a146c0693a5e928983be77a308ad885" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-training-list-500-txt", | |
| "name": "globals-training-list-500-txt", | |
| "description": "Training material ids for the full 500-material set.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/training_list_500.txt", | |
| "encodingFormat": "text/plain", | |
| "contentSize": "1516 B", | |
| "sha256": "921fae5ed3429df0dd1c910294adcd287b529dba266f3c1a86ed9e26f2e4fa99" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-test-list-100-txt", | |
| "name": "globals-test-list-100-txt", | |
| "description": "Test material ids for the 100-material subset.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/test_list_100.txt", | |
| "encodingFormat": "text/plain", | |
| "contentSize": "75 B", | |
| "sha256": "94c42c05331ab0b9b61d80feceb351e7dea5bf44b962a494b6f453626f45383b" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-test-list-300-txt", | |
| "name": "globals-test-list-300-txt", | |
| "description": "Test material ids for the 300-material subset.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/test_list_300.txt", | |
| "encodingFormat": "text/plain", | |
| "contentSize": "225 B", | |
| "sha256": "cf9b7d31e305446238b82635c9da4b63c293b36907ee31847e6061a314d7691b" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-test-list-442-txt", | |
| "name": "globals-test-list-442-txt", | |
| "description": "Test material ids for the 442-material subset.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/test_list_442.txt", | |
| "encodingFormat": "text/plain", | |
| "contentSize": "331 B", | |
| "sha256": "fdd6e41703c405661c85dbf906ba2cc6b17be830ed7e9b10f0ada755bc7976e5" | |
| }, | |
| { | |
| "@type": "cr:FileObject", | |
| "@id": "globals-test-list-500-txt", | |
| "name": "globals-test-list-500-txt", | |
| "description": "Test material ids for the full 500-material set.", | |
| "contentUrl": "https://huggingface.co/datasets/koalapenguin/RoboCloth/resolve/main/globals/test_list_500.txt", | |
| "encodingFormat": "text/plain", | |
| "contentSize": "374 B", | |
| "sha256": "b50fd5522559e311eb5423ae673865ec9ab8e86a237ae829a8c5b27c6eaeb565" | |
| } | |
| ], | |
| "recordSet": [ | |
| { | |
| "@type": "cr:RecordSet", | |
| "@id": "material", | |
| "name": "material", | |
| "description": "One row per cloth material. The material id is extracted from the hdr.tar path and joins to all sibling files at materials/{id}/ in the corresponding FileSets declared in `distribution`.", | |
| "key": { | |
| "@id": "material/id" | |
| }, | |
| "field": [ | |
| { | |
| "@type": "cr:Field", | |
| "@id": "material/id", | |
| "name": "id", | |
| "description": "Numeric material identifier (0-499).", | |
| "dataType": "sc:Integer", | |
| "source": { | |
| "fileSet": { | |
| "@id": "materials-hdr-tar" | |
| }, | |
| "extract": { | |
| "fileProperty": "fullpath" | |
| }, | |
| "transform": { | |
| "regex": "materials/(\\d+)/hdr\\.tar" | |
| } | |
| } | |
| }, | |
| { | |
| "@type": "cr:Field", | |
| "@id": "material/hdr_tar_path", | |
| "name": "hdr_tar_path", | |
| "description": "Repo-relative path to this material's hdr.tar.", | |
| "dataType": "sc:Text", | |
| "source": { | |
| "fileSet": { | |
| "@id": "materials-hdr-tar" | |
| }, | |
| "extract": { | |
| "fileProperty": "fullpath" | |
| } | |
| } | |
| } | |
| ] | |
| } | |
| ], | |
| "sameAs": [ | |
| "https://github.com/theialab/robocloth", | |
| "https://huggingface.co/datasets/koalapenguin/RoboCloth" | |
| ] | |
| } | |