Changes

On December 2, 2024 at 5:39:00 PM UTC, admin:
Set author of MSCOCO to Marc Tanti (previously Jun Yu)
Updated description of MSCOCO from
MSCOCO is a benchmark dataset for various computer vision tasks, including object detection, instance segmentation, and image captioning. It contains 83k training images, 40k validation images, and 81k test images, each associated with five captions.
to
Human Pose Estimation (HPE) aims to estimate the position of each joint point of the human body in a given image. HPE tasks support a wide range of downstream tasks such as activity recognition, motion capture, etc. Recently with the ViT model being proven effective on many visual tasks, many transformer-based methods have achieved excellent performance on HPE tasks.
Added the following tags to MSCOCO
Changed value of field defined_in to https://doi.org/10.48550/arXiv.2403.03715 in MSCOCO
Changed value of field extra_authors to [{'extra_author': 'Albert Gatt', 'orcid': ''}, {'extra_author': 'Adrian Muscat', 'orcid': ''}] in MSCOCO
Changed value of field citation to ['https://doi.org/10.48550/arXiv.1712.02051', 'https://doi.org/10.48550/arXiv.1710.02534', 'https://doi.org/10.48550/arXiv.2304.02828', 'https://doi.org/10.48550/arXiv.2110.01705', 'https://doi.org/10.48550/arXiv.2101.09698', 'https://doi.org/10.48550/arXiv.1704.02798', 'https://doi.org/10.48550/arXiv.2210.12417', 'https://doi.org/10.48550/arXiv.1809.04144', 'https://doi.org/10.48550/arXiv.2405.16759', 'https://doi.org/10.48550/arXiv.2402.18969', 'https://doi.org/10.48550/arXiv.2402.15276', 'https://doi.org/10.48550/arXiv.2008.11073', 'https://doi.org/10.48550/arXiv.1811.04697', 'https://doi.org/10.1007/978-3-030-11018-5_11', 'https://doi.org/10.48550/arXiv.1903.02499', 'https://doi.org/10.1016/j.aiopen.2023.05.001', 'https://doi.org/10.48550/arXiv.1812.00235', 'https://doi.org/10.48550/arXiv.2404.03518', 'https://doi.org/10.48550/arXiv.2010.04963', 'https://doi.org/10.48550/arXiv.2306.07490', 'https://doi.org/10.48550/arXiv.2110.02623', 'https://doi.org/10.48550/arXiv.2006.11807', 'https://doi.org/10.48550/arXiv.2105.03889', 'https://doi.org/10.48550/arXiv.2312.08865', 'https://doi.org/10.48550/arXiv.2206.00621', 'https://doi.org/10.48550/arXiv.1908.04919', 'https://doi.org/10.48550/arXiv.2211.15398', 'https://doi.org/10.48550/arXiv.2310.19654', 'https://doi.org/10.1145/3664647.3681076'] in MSCOCO
Changed value of field landing_page to https://mscoco.org/ in MSCOCO
Deleted resource Original Metadata from MSCOCO
              
    
          
          
        
        
            f 1 { f 1 {
            2   "access_rights": "", 2   "access_rights": "",
            n 3   "author": "Jun Yu", n 3   "author": "Marc Tanti",
            4   "author_email": "", 4   "author_email": "",
            5   "citation": [ 5   "citation": [
            n 6     "https://doi.org/10.18653/v1/D19-5627" n 6     "https://doi.org/10.48550/arXiv.1712.02051",
            7     "https://doi.org/10.48550/arXiv.1710.02534",
            8     "https://doi.org/10.48550/arXiv.2304.02828",
            9     "https://doi.org/10.48550/arXiv.2110.01705",
            10     "https://doi.org/10.48550/arXiv.2101.09698",
            11     "https://doi.org/10.48550/arXiv.1704.02798",
            12     "https://doi.org/10.48550/arXiv.2210.12417",
            13     "https://doi.org/10.48550/arXiv.1809.04144",
            14     "https://doi.org/10.48550/arXiv.2405.16759",
            15     "https://doi.org/10.48550/arXiv.2402.18969",
            16     "https://doi.org/10.48550/arXiv.2402.15276",
            17     "https://doi.org/10.48550/arXiv.2008.11073",
            18     "https://doi.org/10.48550/arXiv.1811.04697",
            19     "https://doi.org/10.1007/978-3-030-11018-5_11",
            20     "https://doi.org/10.48550/arXiv.1903.02499",
            21     "https://doi.org/10.1016/j.aiopen.2023.05.001",
            22     "https://doi.org/10.48550/arXiv.1812.00235",
            23     "https://doi.org/10.48550/arXiv.2404.03518",
            24     "https://doi.org/10.48550/arXiv.2010.04963",
            25     "https://doi.org/10.48550/arXiv.2306.07490",
            26     "https://doi.org/10.48550/arXiv.2110.02623",
            27     "https://doi.org/10.48550/arXiv.2006.11807",
            28     "https://doi.org/10.48550/arXiv.2105.03889",
            29     "https://doi.org/10.48550/arXiv.2312.08865",
            30     "https://doi.org/10.48550/arXiv.2206.00621",
            31     "https://doi.org/10.48550/arXiv.1908.04919",
            32     "https://doi.org/10.48550/arXiv.2211.15398",
            33     "https://doi.org/10.48550/arXiv.2310.19654",
            34     "https://doi.org/10.1145/3664647.3681076"
            7   ], 35   ],
            8   "creator_user_id": "17755db4-395a-4b3b-ac09-e8e3484ca700", 36   "creator_user_id": "17755db4-395a-4b3b-ac09-e8e3484ca700",
            n 9   "defined_in": "https://doi.org/10.48550/arXiv.1905.07841", n 37   "defined_in": "https://doi.org/10.48550/arXiv.2403.03715",
            10   "doi": "10.57702/xriudzva", 38   "doi": "10.57702/xriudzva",
            11   "doi_date_published": "2024-11-25", 39   "doi_date_published": "2024-11-25",
            12   "doi_publisher": "TIB", 40   "doi_publisher": "TIB",
            13   "doi_status": true, 41   "doi_status": true,
            14   "domain": "https://service.tib.eu/ldmservice", 42   "domain": "https://service.tib.eu/ldmservice",
            15   "extra_authors": [ 43   "extra_authors": [
            16     { 44     {
            n 17       "extra_author": "Jing Li", n 45       "extra_author": "Albert Gatt",
            18       "orcid": "" 46       "orcid": ""
            19     }, 47     },
            20     { 48     {
            n 21       "extra_author": "Zhou Yu", n 49       "extra_author": "Adrian Muscat",
            22       "orcid": "" 
            23     }, 
            24     { 
            25       "extra_author": "Qingming Huang", 
            26       "orcid": "" 50       "orcid": ""
            27     } 51     }
            28   ], 52   ],
            n 29   "groups": [], n 53   "groups": [
            54     {
            55       "description": "",
            56       "display_name": "Human Attention",
            57       "id": "03d143ba-0aca-42fc-9bee-d96f37b72948",
            58       "image_display_url": "",
            59       "name": "human-attention",
            60       "title": "Human Attention"
            61     },
            62     {
            63       "description": "",
            64       "display_name": "Human Pose Estimation",
            65       "id": "e880ba2e-3294-47ab-85b0-f66ba44b38bc",
            66       "image_display_url": "",
            67       "name": "human-pose-estimation",
            68       "title": "Human Pose Estimation"
            69     },
            70     {
            71       "description": "",
            72       "display_name": "Image Captioning",
            73       "id": "7a76ce67-2607-4da9-a837-d2017dc33ec6",
            74       "image_display_url": "",
            75       "name": "image-captioning",
            76       "title": "Image Captioning"
            77     },
            78     {
            79       "description": "",
            80       "display_name": "Image Description",
            81       "id": "8d6198ce-1f41-404e-b100-82cb5c33d3d2",
            82       "image_display_url": "",
            83       "name": "image-description",
            84       "title": "Image Description"
            85     },
            86     {
            87       "description": "",
            88       "display_name": "Image Generation",
            89       "id": "be25a76c-def1-4e73-8b1c-b81222d63867",
            90       "image_display_url": "",
            91       "name": "image-generation",
            92       "title": "Image Generation"
            93     },
            94     {
            95       "description": "",
            96       "display_name": "Image-Text Matching",
            97       "id": "0635db3c-45c8-49b1-b083-80ec6966b144",
            98       "image_display_url": "",
            99       "name": "image-text-matching",
            100       "title": "Image-Text Matching"
            101     },
            102     {
            103       "description": "",
            104       "display_name": "Image-Text Retrieval",
            105       "id": "28bbc5c8-e229-4df6-9cf1-d2040f80541a",
            106       "image_display_url": "",
            107       "name": "image-text-retrieval",
            108       "title": "Image-Text Retrieval"
            109     },
            110     {
            111       "description": "",
            112       "display_name": "Images",
            113       "id": "45883730-52c6-44ac-bea8-d50c9cbc8882",
            114       "image_display_url": "",
            115       "name": "images",
            116       "title": "Images"
            117     },
            118     {
            119       "description": "",
            120       "display_name": "Instance Segmentation",
            121       "id": "f856527a-3d35-4c73-8c09-bf3f4a3bbb9f",
            122       "image_display_url": "",
            123       "name": "instance-segmentation",
            124       "title": "Instance Segmentation"
            125     },
            126     {
            127       "description": "",
            128       "display_name": "Multimodal Learning",
            129       "id": "a88cb918-103d-4226-a87d-9d9103c1f839",
            130       "image_display_url": "",
            131       "name": "multimodal-learning",
            132       "title": "Multimodal Learning"
            133     },
            134     {
            135       "description": "",
            136       "display_name": "Object Detection",
            137       "id": "ca2cb1af-d31c-49b0-a1dd-62b22f2b9e20",
            138       "image_display_url": "",
            139       "name": "object-detection",
            140       "title": "Object Detection"
            141     },
            142     {
            143       "description": "",
            144       "display_name": "Paraphrase Generation",
            145       "id": "fdd2258a-e7c3-4fe5-b9bc-5005937d6431",
            146       "image_display_url": "",
            147       "name": "paraphrase-generation",
            148       "title": "Paraphrase Generation"
            149     },
            150     {
            151       "description": "",
            152       "display_name": "Semantic Segmentation",
            153       "id": "8c3f2eee-f5f9-464d-9c0a-1a5e7a925c0e",
            154       "image_display_url": "",
            155       "name": "semantic-segmentation",
            156       "title": "Semantic Segmentation"
            157     },
            158     {
            159       "description": "",
            160       "display_name": "Text-to-Image",
            161       "id": "c68d7763-4eaf-4b60-a763-d8a968e4245e",
            162       "image_display_url": "",
            163       "name": "text-to-image",
            164       "title": "Text-to-Image"
            165     },
            166     {
            167       "description": "",
            168       "display_name": "Vision-and-Language Models",
            169       "id": "8859491a-95fd-495f-9e99-91165571544a",
            170       "image_display_url": "",
            171       "name": "vision-and-language-models",
            172       "title": "Vision-and-Language Models"
            173     },
            174     {
            175       "description": "",
            176       "display_name": "Visual Question Answering",
            177       "id": "e15f82e8-481e-43a8-8a39-d6e0751ace6c",
            178       "image_display_url": "",
            179       "name": "visual-question-answering",
            180       "title": "Visual Question Answering"
            181     }
            182   ],
            30   "id": "b55ff6e8-0c5e-49c7-9281-f8b174aa09f3", 183   "id": "b55ff6e8-0c5e-49c7-9281-f8b174aa09f3",
            31   "isopen": false, 184   "isopen": false,
            n 32   "landing_page": "https://cocodataset.org/", n 185   "landing_page": "https://mscoco.org/",
            33   "license_title": null, 186   "license_title": null,
            34   "link_orkg": "", 187   "link_orkg": "",
            35   "metadata_created": "2024-11-25T14:18:37.775582", 188   "metadata_created": "2024-11-25T14:18:37.775582",
            n 36   "metadata_modified": "2024-11-25T14:31:47.455333", n 189   "metadata_modified": "2024-12-02T17:38:59.395188",
            37   "name": "mscoco", 190   "name": "mscoco",
            n 38   "notes": "MSCOCO is a benchmark dataset for various computer vision  n 191   "notes": "Human Pose Estimation (HPE) aims to estimate the position 
            39 tasks, including object detection, instance segmentation, and image  192 of each joint point of the human body in a given image. HPE tasks 
            40 captioning. It contains 83k training images, 40k validation images,  193 support a wide range of downstream tasks such as activity recognition, 
            41 and 81k test images, each associated with five captions.", 194 motion capture, etc. Recently with the ViT model being proven 
            195 effective on many visual tasks, many transformer-based methods have 
            196 achieved excellent performance on HPE tasks.",
            42   "num_resources": 1, 197   "num_resources": 0,
            43   "num_tags": 0, 198   "num_tags": 49,
            44   "organization": { 199   "organization": {
            45     "approval_status": "approved", 200     "approval_status": "approved",
            46     "created": "2024-11-25T12:11:38.292601", 201     "created": "2024-11-25T12:11:38.292601",
            47     "description": "", 202     "description": "",
            48     "id": "079d46db-32df-4b48-91f3-0a8bc8f69559", 203     "id": "079d46db-32df-4b48-91f3-0a8bc8f69559",
            49     "image_url": "", 204     "image_url": "",
            50     "is_organization": true, 205     "is_organization": true,
            51     "name": "no-organization", 206     "name": "no-organization",
            52     "state": "active", 207     "state": "active",
            53     "title": "No Organization", 208     "title": "No Organization",
            54     "type": "organization" 209     "type": "organization"
            55   }, 210   },
            56   "owner_org": "079d46db-32df-4b48-91f3-0a8bc8f69559", 211   "owner_org": "079d46db-32df-4b48-91f3-0a8bc8f69559",
            57   "private": false, 212   "private": false,
            58   "relationships_as_object": [], 213   "relationships_as_object": [],
            59   "relationships_as_subject": [], 214   "relationships_as_subject": [],
            n 60   "resources": [ n 215   "resources": [],
            61     { 216   "services_used_list": "",
            62       "cache_last_updated": null, 
            63       "cache_url": null, 
            64       "created": "2024-11-25T15:30:26", 
            65       "data": [ 
            66         "dcterms:title", 
            67         "dcterms:accessRights", 
            68         "dcterms:creator", 
            69         "dcterms:description", 
            70         "dcterms:issued", 
            71         "dcterms:language", 
            72         "dcterms:identifier", 
            73         "dcat:theme", 
            74         "dcterms:type", 
            75         "dcat:keyword", 
            76         "dcat:landingPage", 
            77         "dcterms:hasVersion", 
            78         "dcterms:format", 
            79         "mls:task", 
            80         "datacite:isDescribedBy" 
            81       ], 
            82       "description": "The json representation of the dataset with its  
            83 distributions based on DCAT.", 
            84       "format": "JSON", 
            85       "hash": "", 
            86       "id": "df12f624-61cf-441a-af25-2d04f3fec39a", 
            87       "last_modified": "2024-11-25T14:31:47.447621", 
            88       "metadata_modified": "2024-11-25T14:31:47.458648", 
            89       "mimetype": "application/json", 
            90       "mimetype_inner": null, 
            91       "name": "Original Metadata", 
            92       "package_id": "b55ff6e8-0c5e-49c7-9281-f8b174aa09f3", 
            93       "position": 0, 
            94       "resource_type": null, 
            95       "size": 1076, 
            96       "state": "active", 217   "state": "active",
            97       "url":  218   "tags": [
            98 resource/df12f624-61cf-441a-af25-2d04f3fec39a/download/metadata.json", 219     {
            99       "url_type": "upload" 220       "display_name": "Captioning",
            221       "id": "99ec7549-b23a-432e-88f4-803dd8685b2c",
            222       "name": "Captioning",
            223       "state": "active",
            224       "vocabulary_id": null
            225     },
            226     {
            227       "display_name": "Image Captioning",
            228       "id": "c708cba4-0a1f-45c9-826f-898857783343",
            229       "name": "Image Captioning",
            230       "state": "active",
            231       "vocabulary_id": null
            232     },
            233     {
            234       "display_name": "Image Description",
            235       "id": "d3ef3b7e-c8b4-4a5e-be62-11eed53a5567",
            236       "name": "Image Description",
            237       "state": "active",
            238       "vocabulary_id": null
            239     },
            240     {
            241       "display_name": "Image Retrieval",
            242       "id": "4b81e5bb-1f8f-415b-a688-0e2d94b6d499",
            243       "name": "Image Retrieval",
            244       "state": "active",
            245       "vocabulary_id": null
            246     },
            247     {
            248       "display_name": "Image-Text Matching",
            249       "id": "84540152-adb0-4565-a5c1-cf49882b8caa",
            250       "name": "Image-Text Matching",
            251       "state": "active",
            252       "vocabulary_id": null
            253     },
            254     {
            255       "display_name": "Images",
            256       "id": "b653d5cb-88da-4373-ae34-3b0b41d37fad",
            257       "name": "Images",
            258       "state": "active",
            259       "vocabulary_id": null
            260     },
            261     {
            262       "display_name": "Instance Segmentation",
            263       "id": "b58d8dfe-1216-401d-8a2a-ceb09e07a013",
            264       "name": "Instance Segmentation",
            265       "state": "active",
            266       "vocabulary_id": null
            267     },
            268     {
            269       "display_name": "Japanese",
            270       "id": "e7a703ef-c107-49e5-b9c8-5b9a109cf39b",
            271       "name": "Japanese",
            272       "state": "active",
            273       "vocabulary_id": null
            274     },
            275     {
            276       "display_name": "MSCOCO",
            277       "id": "1405508e-1059-4b81-93e7-bdb979347d50",
            278       "name": "MSCOCO",
            279       "state": "active",
            280       "vocabulary_id": null
            281     },
            282     {
            283       "display_name": "Multimodal Learning",
            284       "id": "41329983-d90e-4560-a97d-7fa2ba0a1f92",
            285       "name": "Multimodal Learning",
            286       "state": "active",
            287       "vocabulary_id": null
            288     },
            289     {
            290       "display_name": "Object Detection",
            291       "id": "44adc011-570b-46cf-9a65-ab72ca690477",
            292       "name": "Object Detection",
            293       "state": "active",
            294       "vocabulary_id": null
            295     },
            296     {
            297       "display_name": "Semantic Segmentation",
            298       "id": "809ad6af-28cd-43bd-974d-055a5c0f2973",
            299       "name": "Semantic Segmentation",
            300       "state": "active",
            301       "vocabulary_id": null
            302     },
            303     {
            304       "display_name": "Visual Question Answering",
            305       "id": "d8aee55f-7e04-411f-8ce5-0794bc0406b0",
            306       "name": "Visual Question Answering",
            307       "state": "active",
            308       "vocabulary_id": null
            309     },
            310     {
            311       "display_name": "Weakly Supervised Learning",
            312       "id": "3bf0db7d-8b83-4928-bc08-62c0e97f7dfd",
            313       "name": "Weakly Supervised Learning",
            314       "state": "active",
            315       "vocabulary_id": null
            316     },
            317     {
            318       "display_name": "computer vision",
            319       "id": "f650b4e3-9955-49b0-ba7b-2d302a990978",
            320       "name": "computer vision",
            321       "state": "active",
            322       "vocabulary_id": null
            323     },
            324     {
            325       "display_name": "dataset",
            326       "id": "ce5ad030-ca3d-47e6-abd1-5c92a2806f1b",
            327       "name": "dataset",
            328       "state": "active",
            329       "vocabulary_id": null
            330     },
            331     {
            332       "display_name": "diffusion models",
            333       "id": "67686580-d41a-4e00-875a-494e99951342",
            334       "name": "diffusion models",
            335       "state": "active",
            336       "vocabulary_id": null
            337     },
            338     {
            339       "display_name": "gender",
            340       "id": "af90adbe-d62d-4309-ba7b-132696a7db64",
            341       "name": "gender",
            342       "state": "active",
            343       "vocabulary_id": null
            344     },
            345     {
            346       "display_name": "human attention",
            347       "id": "a4f7720f-dbdc-4830-a369-dacf1fd59bc9",
            348       "name": "human attention",
            349       "state": "active",
            350       "vocabulary_id": null
            351     },
            352     {
            353       "display_name": "human pose estimation",
            354       "id": "2ba9bb7a-941e-4f02-8376-06b50691a7dd",
            355       "name": "human pose estimation",
            356       "state": "active",
            357       "vocabulary_id": null
            358     },
            359     {
            360       "display_name": "image captioning",
            361       "id": "f1bbe827-a03a-4280-b9fa-0599ccfc0541",
            362       "name": "image captioning",
            363       "state": "active",
            364       "vocabulary_id": null
            365     },
            366     {
            367       "display_name": "image classification",
            368       "id": "34936550-ce1a-41b5-8c58-23081a6c673d",
            369       "name": "image classification",
            370       "state": "active",
            371       "vocabulary_id": null
            372     },
            373     {
            374       "display_name": "image dataset",
            375       "id": "d3acafab-ad07-46a1-88d5-540c2fd41466",
            376       "name": "image dataset",
            377       "state": "active",
            378       "vocabulary_id": null
            379     },
            380     {
            381       "display_name": "image description",
            382       "id": "99e3a7d7-7ee7-4f66-adf4-28ac19f60923",
            383       "name": "image description",
            384       "state": "active",
            385       "vocabulary_id": null
            386     },
            387     {
            388       "display_name": "image generation",
            389       "id": "96df81b4-32fd-4826-a903-affb005a0a60",
            390       "name": "image generation",
            391       "state": "active",
            392       "vocabulary_id": null
            393     },
            394     {
            395       "display_name": "image processing",
            396       "id": "12b6291b-2e80-4736-935f-cc942880df77",
            397       "name": "image processing",
            398       "state": "active",
            399       "vocabulary_id": null
            400     },
            401     {
            402       "display_name": "image-text matching",
            403       "id": "75c56b7d-ab6b-4dd8-88f8-67e92928dca0",
            404       "name": "image-text matching",
            405       "state": "active",
            406       "vocabulary_id": null
            407     },
            408     {
            409       "display_name": "image-text pairs",
            410       "id": "07cb3691-65c6-4b58-b3bf-5290980a86e3",
            411       "name": "image-text pairs",
            412       "state": "active",
            413       "vocabulary_id": null
            414     },
            415     {
            416       "display_name": "image-text retrieval",
            417       "id": "2ce62fa7-a31e-4be3-9312-92be69a8726b",
            418       "name": "image-text retrieval",
            419       "state": "active",
            420       "vocabulary_id": null
            421     },
            422     {
            423       "display_name": "key-point based object detection",
            424       "id": "510e6689-8d4d-4eee-9fbb-6c0d8f8c9ae9",
            425       "name": "key-point based object detection",
            426       "state": "active",
            427       "vocabulary_id": null
            428     },
            429     {
            430       "display_name": "large-scale dataset",
            431       "id": "a9c694bf-f591-4625-a20e-d53d3f90d489",
            432       "name": "large-scale dataset",
            433       "state": "active",
            434       "vocabulary_id": null
            435     },
            436     {
            437       "display_name": "natural descriptions",
            438       "id": "cc6fad82-69e6-4d60-bf91-fcb10fc975e1",
            439       "name": "natural descriptions",
            440       "state": "active",
            441       "vocabulary_id": null
            442     },
            443     {
            444       "display_name": "natural language processing",
            445       "id": "8af9c93a-1d87-41e0-83d9-f5d01a2bbd0c",
            446       "name": "natural language processing",
            447       "state": "active",
            448       "vocabulary_id": null
            449     },
            450     {
            451       "display_name": "neural networks",
            452       "id": "8ef5d32b-4073-4531-87bc-4b4676fd22b8",
            453       "name": "neural networks",
            454       "state": "active",
            455       "vocabulary_id": null
            456     },
            457     {
            458       "display_name": "non-autoregressive",
            459       "id": "a65d259a-0ed2-4b9f-b3cd-fc666dd6c6a5",
            460       "name": "non-autoregressive",
            461       "state": "active",
            462       "vocabulary_id": null
            463     },
            464     {
            465       "display_name": "object detection",
            466       "id": "607283c7-9e12-4167-9101-7f8078fb6537",
            467       "name": "object detection",
            468       "state": "active",
            469       "vocabulary_id": null
            470     },
            471     {
            472       "display_name": "paraphrase generation",
            473       "id": "32754920-b11e-4892-b082-84ade55232ba",
            474       "name": "paraphrase generation",
            475       "state": "active",
            476       "vocabulary_id": null
            477     },
            478     {
            479       "display_name": "sequence generation",
            480       "id": "92c922a0-5cdc-48db-be04-ea01a0fba058",
            481       "name": "sequence generation",
            482       "state": "active",
            483       "vocabulary_id": null
            484     },
            485     {
            486       "display_name": "single-shot detectors",
            487       "id": "714b183c-81a8-4198-bbd1-46ed5cda6f29",
            488       "name": "single-shot detectors",
            489       "state": "active",
            490       "vocabulary_id": null
            491     },
            492     {
            493       "display_name": "skin-tone",
            494       "id": "6687fb65-1ff2-43c4-944f-ea24297837f4",
            495       "name": "skin-tone",
            496       "state": "active",
            497       "vocabulary_id": null
            498     },
            499     {
            500       "display_name": "synthetic pairs",
            501       "id": "1aa7c615-584c-4798-8b2f-00cdfe43ae84",
            502       "name": "synthetic pairs",
            503       "state": "active",
            504       "vocabulary_id": null
            505     },
            506     {
            507       "display_name": "text-only",
            508       "id": "ef11147f-7893-48cd-adf6-414d9f971718",
            509       "name": "text-only",
            510       "state": "active",
            511       "vocabulary_id": null
            512     },
            513     {
            514       "display_name": "text-to-image",
            515       "id": "0b0e2f9a-321b-4b04-abc6-14130f66400c",
            516       "name": "text-to-image",
            517       "state": "active",
            518       "vocabulary_id": null
            519     },
            520     {
            521       "display_name": "vision-and-language models",
            522       "id": "7645deef-42d9-4294-a894-9829f9c04122",
            523       "name": "vision-and-language models",
            524       "state": "active",
            525       "vocabulary_id": null
            526     },
            527     {
            528       "display_name": "vision-language pre-training",
            529       "id": "fd01e9f9-7fca-46ce-83d8-ff81c76d8e7a",
            530       "name": "vision-language pre-training",
            531       "state": "active",
            532       "vocabulary_id": null
            533     },
            534     {
            535       "display_name": "visual language grounding",
            536       "id": "d2b22694-0a09-4953-9275-2ed27bfbab03",
            537       "name": "visual language grounding",
            538       "state": "active",
            539       "vocabulary_id": null
            540     },
            541     {
            542       "display_name": "visual question answering",
            543       "id": "8be24887-e029-464f-b56c-1d0a460e2005",
            544       "name": "visual question answering",
            545       "state": "active",
            546       "vocabulary_id": null
            547     },
            548     {
            549       "display_name": "visual relationship detection",
            550       "id": "e1f83c99-b862-498b-b955-ee73b1648e6f",
            551       "name": "visual relationship detection",
            552       "state": "active",
            553       "vocabulary_id": null
            554     },
            555     {
            556       "display_name": "visual-textual embedding",
            557       "id": "6ee32e99-b284-45d5-aeaf-59560d8ef3bb",
            558       "name": "visual-textual embedding",
            559       "state": "active",
            560       "vocabulary_id": null
            100     } 561     }
            101   ], 562   ],
            t 102   "services_used_list": "", t 
            103   "state": "active", 
            104   "tags": [], 
            105   "title": "MSCOCO", 563   "title": "MSCOCO",
            106   "type": "dataset", 564   "type": "dataset",
            107   "version": "" 565   "version": ""
            108 } 566 }