Changes

On December 2, 2024 at 5:52:17 PM UTC, admin:
Set author of Visual Genome to Ranjay Krishna (previously Zhou Yu)
Updated description of Visual Genome from
Visual Genome is a large-scale dataset that provides dense image annotations with object, attribute, and relationship annotations. It contains 108k images and is utilized to pre-train the object detectors in the study.
to
The Visual Genome dataset is a large-scale visual question answering dataset, containing 1.5 million images, each with 15-30 annotated entities, attributes, and relationships.
Added the following tags to Visual Genome
Changed value of field defined_in to https://doi.org/10.48550/arXiv.2304.14573 in Visual Genome
Changed value of field extra_authors to [{'extra_author': 'Yuke Zhu', 'orcid': ''}, {'extra_author': 'Oliver Groth', 'orcid': ''}, {'extra_author': 'Justin Johnson', 'orcid': ''}, {'extra_author': 'Kenji Hata', 'orcid': ''}, {'extra_author': 'Joshua Kravitz', 'orcid': ''}, {'extra_author': 'Stephanie Chen', 'orcid': ''}, {'extra_author': 'Yannis Kalantidis', 'orcid': ''}, {'extra_author': 'Li-Jia Li', 'orcid': ''}, {'extra_author': 'David A Shamma', 'orcid': ''}] in Visual Genome
Changed value of field citation to ['https://doi.org/10.48550/arXiv.1711.08389', 'https://doi.org/10.48550/arXiv.1812.02378', 'https://doi.org/10.48550/arXiv.2404.00906', 'https://doi.org/10.48550/arXiv.2006.03776', 'https://doi.org/10.48550/arXiv.2003.07012', 'https://doi.org/10.48550/arXiv.1902.05715', 'https://doi.org/10.48550/arXiv.2307.04838', 'https://doi.org/10.48550/arXiv.2008.12679', 'https://doi.org/10.48550/arXiv.1906.03952', 'https://doi.org/10.48550/arXiv.2211.16636', 'https://doi.org/10.48550/arXiv.1904.01410', 'https://doi.org/10.48550/arXiv.1803.08896', 'https://doi.org/10.48550/arXiv.2405.15321', 'https://doi.org/10.48550/arXiv.2206.08358', 'https://doi.org/10.48550/arXiv.1811.11903', 'https://doi.org/10.48550/arXiv.2212.00280', 'https://doi.org/10.48550/arXiv.2201.06696', 'https://doi.org/10.48550/arXiv.1909.09953', 'https://doi.org/10.48550/arXiv.2007.04422', 'https://doi.org/10.48550/arXiv.1801.05096', 'https://doi.org/10.48550/arXiv.2305.12476', 'https://doi.org/10.48550/arXiv.1708.06819', 'https://doi.org/10.48550/arXiv.2208.09374', 'https://doi.org/10.48550/arXiv.2206.08919', 'https://doi.org/10.48550/arXiv.2208.04060', 'https://doi.org/10.48550/arXiv.2201.09880', 'https://doi.org/10.48550/arXiv.2210.15327', 'https://doi.org/10.48550/arXiv.2205.00363', 'https://doi.org/10.48550/arXiv.1903.05434', 'https://doi.org/10.48550/arXiv.2006.11807', 'https://doi.org/10.48550/arXiv.2102.04990', 'https://doi.org/10.48550/arXiv.2211.15398'] in Visual Genome
Changed value of field landing_page to https://visualgenome.org/ in Visual Genome
Deleted resource Original Metadata from Visual Genome
              
    
          
          
        
        
            f 1 { f 1 {
            2   "access_rights": "", 2   "access_rights": "",
            n 3   "author": "Zhou Yu", n 3   "author": "Ranjay Krishna",
            4   "author_email": "", 4   "author_email": "",
            5   "citation": [ 5   "citation": [
            n n 6     "https://doi.org/10.48550/arXiv.1711.08389",
            6     "https://doi.org/10.48550/arXiv.1912.12394", 7     "https://doi.org/10.48550/arXiv.1812.02378",
            8     "https://doi.org/10.48550/arXiv.2404.00906",
            9     "https://doi.org/10.48550/arXiv.2006.03776",
            10     "https://doi.org/10.48550/arXiv.2003.07012",
            7     "https://doi.org/10.48550/arXiv.1905.07841" 11     "https://doi.org/10.48550/arXiv.1902.05715",
            12     "https://doi.org/10.48550/arXiv.2307.04838",
            13     "https://doi.org/10.48550/arXiv.2008.12679",
            14     "https://doi.org/10.48550/arXiv.1906.03952",
            15     "https://doi.org/10.48550/arXiv.2211.16636",
            16     "https://doi.org/10.48550/arXiv.1904.01410",
            17     "https://doi.org/10.48550/arXiv.1803.08896",
            18     "https://doi.org/10.48550/arXiv.2405.15321",
            19     "https://doi.org/10.48550/arXiv.2206.08358",
            20     "https://doi.org/10.48550/arXiv.1811.11903",
            21     "https://doi.org/10.48550/arXiv.2212.00280",
            22     "https://doi.org/10.48550/arXiv.2201.06696",
            23     "https://doi.org/10.48550/arXiv.1909.09953",
            24     "https://doi.org/10.48550/arXiv.2007.04422",
            25     "https://doi.org/10.48550/arXiv.1801.05096",
            26     "https://doi.org/10.48550/arXiv.2305.12476",
            27     "https://doi.org/10.48550/arXiv.1708.06819",
            28     "https://doi.org/10.48550/arXiv.2208.09374",
            29     "https://doi.org/10.48550/arXiv.2206.08919",
            30     "https://doi.org/10.48550/arXiv.2208.04060",
            31     "https://doi.org/10.48550/arXiv.2201.09880",
            32     "https://doi.org/10.48550/arXiv.2210.15327",
            33     "https://doi.org/10.48550/arXiv.2205.00363",
            34     "https://doi.org/10.48550/arXiv.1903.05434",
            35     "https://doi.org/10.48550/arXiv.2006.11807",
            36     "https://doi.org/10.48550/arXiv.2102.04990",
            37     "https://doi.org/10.48550/arXiv.2211.15398"
            8   ], 38   ],
            9   "creator_user_id": "17755db4-395a-4b3b-ac09-e8e3484ca700", 39   "creator_user_id": "17755db4-395a-4b3b-ac09-e8e3484ca700",
            n 10   "defined_in": "https://doi.org/10.48550/arXiv.2002.07417", n 40   "defined_in": "https://doi.org/10.48550/arXiv.2304.14573",
            11   "doi": "10.57702/p2euk5d9", 41   "doi": "10.57702/p2euk5d9",
            12   "doi_date_published": "2024-11-25", 42   "doi_date_published": "2024-11-25",
            13   "doi_publisher": "TIB", 43   "doi_publisher": "TIB",
            14   "doi_status": true, 44   "doi_status": true,
            15   "domain": "https://service.tib.eu/ldmservice", 45   "domain": "https://service.tib.eu/ldmservice",
            16   "extra_authors": [ 46   "extra_authors": [
            17     { 47     {
            n n 48       "extra_author": "Yuke Zhu",
            49       "orcid": ""
            50     },
            51     {
            52       "extra_author": "Oliver Groth",
            53       "orcid": ""
            54     },
            55     {
            56       "extra_author": "Justin Johnson",
            57       "orcid": ""
            58     },
            59     {
            18       "extra_author": "Qingming Huang", 60       "extra_author": "Kenji Hata",
            61       "orcid": ""
            62     },
            63     {
            64       "extra_author": "Joshua Kravitz",
            65       "orcid": ""
            66     },
            67     {
            68       "extra_author": "Stephanie Chen",
            69       "orcid": ""
            70     },
            71     {
            72       "extra_author": "Yannis Kalantidis",
            73       "orcid": ""
            74     },
            75     {
            76       "extra_author": "Li-Jia Li",
            77       "orcid": ""
            78     },
            79     {
            80       "extra_author": "David A Shamma",
            19       "orcid": "" 81       "orcid": ""
            20     } 82     }
            21   ], 83   ],
            n 22   "groups": [], n 84   "groups": [
            85     {
            86       "description": "",
            87       "display_name": "Image Analysis",
            88       "id": "41147ff5-6a94-463c-9fe3-83db1dbf84a8",
            89       "image_display_url": "",
            90       "name": "image-analysis",
            91       "title": "Image Analysis"
            92     },
            93     {
            94       "description": "",
            95       "display_name": "Image Annotation",
            96       "id": "0d7d1f08-f908-466b-a978-2ea0573fe19f",
            97       "image_display_url": "",
            98       "name": "image-annotation",
            99       "title": "Image Annotation"
            100     },
            101     {
            102       "description": "",
            103       "display_name": "Image Captioning",
            104       "id": "7a76ce67-2607-4da9-a837-d2017dc33ec6",
            105       "image_display_url": "",
            106       "name": "image-captioning",
            107       "title": "Image Captioning"
            108     },
            109     {
            110       "description": "",
            111       "display_name": "Image Dataset",
            112       "id": "fc745cca-b21e-4ced-ba81-06a456938edf",
            113       "image_display_url": "",
            114       "name": "image-dataset",
            115       "title": "Image Dataset"
            116     },
            117     {
            118       "description": "",
            119       "display_name": "Image Description",
            120       "id": "8d6198ce-1f41-404e-b100-82cb5c33d3d2",
            121       "image_display_url": "",
            122       "name": "image-description",
            123       "title": "Image Description"
            124     },
            125     {
            126       "description": "",
            127       "display_name": "Image Segmentation",
            128       "id": "7c8cc5f1-a9b2-4924-82ec-9e3aa3049a04",
            129       "image_display_url": "",
            130       "name": "image-segmentation",
            131       "title": "Image Segmentation"
            132     },
            133     {
            134       "description": "",
            135       "display_name": "Image Synthesis",
            136       "id": "8b89ca6b-96f7-439b-8045-febd38230620",
            137       "image_display_url": "",
            138       "name": "image-synthesis",
            139       "title": "Image Synthesis"
            140     },
            141     {
            142       "description": "",
            143       "display_name": "Image Understanding",
            144       "id": "638a25dd-1bfb-497e-89d2-287a377aa4f6",
            145       "image_display_url": "",
            146       "name": "image-understanding",
            147       "title": "Image Understanding"
            148     },
            149     {
            150       "description": "",
            151       "display_name": "Image-Text Matching",
            152       "id": "0635db3c-45c8-49b1-b083-80ec6966b144",
            153       "image_display_url": "",
            154       "name": "image-text-matching",
            155       "title": "Image-Text Matching"
            156     },
            157     {
            158       "description": "",
            159       "display_name": "Image-Text Retrieval",
            160       "id": "28bbc5c8-e229-4df6-9cf1-d2040f80541a",
            161       "image_display_url": "",
            162       "name": "image-text-retrieval",
            163       "title": "Image-Text Retrieval"
            164     },
            165     {
            166       "description": "",
            167       "display_name": "Image-to-Text Embedding",
            168       "id": "bdf50d0c-b367-43bc-8875-a552cc636bcb",
            169       "image_display_url": "",
            170       "name": "image-to-text-embedding",
            171       "title": "Image-to-Text Embedding"
            172     },
            173     {
            174       "description": "",
            175       "display_name": "Natural Language Processing",
            176       "id": "5b974bcc-8f79-40fc-a05d-75b861d55ed0",
            177       "image_display_url": "",
            178       "name": "natural-language-processing",
            179       "title": "Natural Language Processing"
            180     },
            181     {
            182       "description": "",
            183       "display_name": "Object Detection",
            184       "id": "ca2cb1af-d31c-49b0-a1dd-62b22f2b9e20",
            185       "image_display_url": "",
            186       "name": "object-detection",
            187       "title": "Object Detection"
            188     },
            189     {
            190       "description": "",
            191       "display_name": "Question Answering",
            192       "id": "01e9bd6b-910d-421d-bb31-799604023211",
            193       "image_display_url": "",
            194       "name": "question-answering",
            195       "title": "Question Answering"
            196     },
            197     {
            198       "description": "",
            199       "display_name": "Scene Graph Generation",
            200       "id": "8aa99886-50d9-408e-9591-c6545536ccb4",
            201       "image_display_url": "",
            202       "name": "scene-graph-generation",
            203       "title": "Scene Graph Generation"
            204     },
            205     {
            206       "description": "",
            207       "display_name": "Scene Graphs",
            208       "id": "17df02e9-e303-483c-b1cd-c3c61612e220",
            209       "image_display_url": "",
            210       "name": "scene-graphs",
            211       "title": "Scene Graphs"
            212     },
            213     {
            214       "description": "",
            215       "display_name": "Scene Understanding",
            216       "id": "f3a9e19d-4b53-4b7d-bd4b-65c69132188a",
            217       "image_display_url": "",
            218       "name": "scene-understanding",
            219       "title": "Scene Understanding"
            220     },
            221     {
            222       "description": "",
            223       "display_name": "Vision-Language Pre-training",
            224       "id": "58ef4f39-fe42-4d9a-9f40-3951b2551069",
            225       "image_display_url": "",
            226       "name": "vision-language-pre-training",
            227       "title": "Vision-Language Pre-training"
            228     },
            229     {
            230       "description": "",
            231       "display_name": "Visual Grounding",
            232       "id": "0ecf0e36-85b5-40e9-9ca0-80a21e99461b",
            233       "image_display_url": "",
            234       "name": "visual-grounding",
            235       "title": "Visual Grounding"
            236     },
            237     {
            238       "description": "",
            239       "display_name": "Visual Question Answering",
            240       "id": "e15f82e8-481e-43a8-8a39-d6e0751ace6c",
            241       "image_display_url": "",
            242       "name": "visual-question-answering",
            243       "title": "Visual Question Answering"
            244     },
            245     {
            246       "description": "",
            247       "display_name": "Visual Reasoning",
            248       "id": "a94d0db9-a53a-46be-a7d9-4b5e103342fd",
            249       "image_display_url": "",
            250       "name": "visual-reasoning",
            251       "title": "Visual Reasoning"
            252     },
            253     {
            254       "description": "",
            255       "display_name": "Visual Relation Detection",
            256       "id": "ce3d615e-993a-4870-bc23-1b03a3fe93c4",
            257       "image_display_url": "",
            258       "name": "visual-relation-detection",
            259       "title": "Visual Relation Detection"
            260     },
            261     {
            262       "description": "",
            263       "display_name": "Visual Relationship Detection",
            264       "id": "0f963144-4535-464d-9bf2-7d7ae4a5422b",
            265       "image_display_url": "",
            266       "name": "visual-relationship-detection",
            267       "title": "Visual Relationship Detection"
            268     },
            269     {
            270       "description": "",
            271       "display_name": "Visual Relationship Prediction",
            272       "id": "8bf3722b-3cf1-420d-bd25-1da61f68a2ac",
            273       "image_display_url": "",
            274       "name": "visual-relationship-prediction",
            275       "title": "Visual Relationship Prediction"
            276     },
            277     {
            278       "description": "",
            279       "display_name": "Visual Understanding",
            280       "id": "17164606-559d-4fa7-a1c6-4624298074e1",
            281       "image_display_url": "",
            282       "name": "visual-understanding",
            283       "title": "Visual Understanding"
            284     }
            285   ],
            23   "id": "9f7d7ea8-8aba-4e09-8bad-da5a88d28558", 286   "id": "9f7d7ea8-8aba-4e09-8bad-da5a88d28558",
            24   "isopen": false, 287   "isopen": false,
            n 25   "landing_page": "http://visualgenome.org/", n 288   "landing_page": "https://visualgenome.org/",
            26   "license_title": null, 289   "license_title": null,
            27   "link_orkg": "", 290   "link_orkg": "",
            28   "metadata_created": "2024-11-25T14:18:39.433673", 291   "metadata_created": "2024-11-25T14:18:39.433673",
            n 29   "metadata_modified": "2024-11-25T14:31:49.268551", n 292   "metadata_modified": "2024-12-02T17:52:16.593230",
            30   "name": "visual-genome", 293   "name": "visual-genome",
            n 31   "notes": "Visual Genome is a large-scale dataset that provides dense  n 294   "notes": "The Visual Genome dataset is a large-scale visual question 
            32 image annotations with object, attribute, and relationship  295 answering dataset, containing 1.5 million images, each with 15-30 
            33 annotations. It contains 108k images and is utilized to pre-train the  296 annotated entities, attributes, and relationships.",
            34 object detectors in the study.", 
            35   "num_resources": 1, 297   "num_resources": 0,
            36   "num_tags": 0, 298   "num_tags": 58,
            37   "organization": { 299   "organization": {
            38     "approval_status": "approved", 300     "approval_status": "approved",
            39     "created": "2024-11-25T12:11:38.292601", 301     "created": "2024-11-25T12:11:38.292601",
            40     "description": "", 302     "description": "",
            41     "id": "079d46db-32df-4b48-91f3-0a8bc8f69559", 303     "id": "079d46db-32df-4b48-91f3-0a8bc8f69559",
            42     "image_url": "", 304     "image_url": "",
            43     "is_organization": true, 305     "is_organization": true,
            44     "name": "no-organization", 306     "name": "no-organization",
            45     "state": "active", 307     "state": "active",
            46     "title": "No Organization", 308     "title": "No Organization",
            47     "type": "organization" 309     "type": "organization"
            48   }, 310   },
            49   "owner_org": "079d46db-32df-4b48-91f3-0a8bc8f69559", 311   "owner_org": "079d46db-32df-4b48-91f3-0a8bc8f69559",
            50   "private": false, 312   "private": false,
            51   "relationships_as_object": [], 313   "relationships_as_object": [],
            52   "relationships_as_subject": [], 314   "relationships_as_subject": [],
            n 53   "resources": [ n 315   "resources": [],
            54     { 316   "services_used_list": "",
            55       "cache_last_updated": null, 
            56       "cache_url": null, 
            57       "created": "2024-11-25T15:30:26", 
            58       "data": [ 
            59         "dcterms:title", 
            60         "dcterms:accessRights", 
            61         "dcterms:creator", 
            62         "dcterms:description", 
            63         "dcterms:issued", 
            64         "dcterms:language", 
            65         "dcterms:identifier", 
            66         "dcat:theme", 
            67         "dcterms:type", 
            68         "dcat:keyword", 
            69         "dcat:landingPage", 
            70         "dcterms:hasVersion", 
            71         "dcterms:format", 
            72         "mls:task", 
            73         "datacite:isDescribedBy" 
            74       ], 
            75       "description": "The json representation of the dataset with its  
            76 distributions based on DCAT.", 
            77       "format": "JSON", 
            78       "hash": "", 
            79       "id": "fbb447e7-385f-45f9-b638-fdc642315cb6", 
            80       "last_modified": "2024-11-25T14:31:49.261215", 
            81       "metadata_modified": "2024-11-25T14:31:49.272186", 
            82       "mimetype": "application/json", 
            83       "mimetype_inner": null, 
            84       "name": "Original Metadata", 
            85       "package_id": "9f7d7ea8-8aba-4e09-8bad-da5a88d28558", 
            86       "position": 0, 
            87       "resource_type": null, 
            88       "size": 1127, 
            89       "state": "active", 317   "state": "active",
            90       "url":  318   "tags": [
            91 resource/fbb447e7-385f-45f9-b638-fdc642315cb6/download/metadata.json", 319     {
            92       "url_type": "upload" 320       "display_name": "Dataset",
            321       "id": "81587eb2-9569-4a4b-83c8-0e2ac78e7e3b",
            322       "name": "Dataset",
            323       "state": "active",
            324       "vocabulary_id": null
            325     },
            326     {
            327       "display_name": "Image",
            328       "id": "ca701146-4139-43fe-a114-4f31ab3d20a1",
            329       "name": "Image",
            330       "state": "active",
            331       "vocabulary_id": null
            332     },
            333     {
            334       "display_name": "Image Analysis",
            335       "id": "47180f4c-083f-4e73-bf21-2cea322a428c",
            336       "name": "Image Analysis",
            337       "state": "active",
            338       "vocabulary_id": null
            339     },
            340     {
            341       "display_name": "Image annotation",
            342       "id": "80baf43e-0465-4d96-9ecf-8aeef2acfada",
            343       "name": "Image annotation",
            344       "state": "active",
            345       "vocabulary_id": null
            346     },
            347     {
            348       "display_name": "Image-Text Matching",
            349       "id": "84540152-adb0-4565-a5c1-cf49882b8caa",
            350       "name": "Image-Text Matching",
            351       "state": "active",
            352       "vocabulary_id": null
            353     },
            354     {
            355       "display_name": "Large Scale Dataset",
            356       "id": "feb463dc-37cd-4b00-a5ee-dde0ed3804bf",
            357       "name": "Large Scale Dataset",
            358       "state": "active",
            359       "vocabulary_id": null
            360     },
            361     {
            362       "display_name": "Object Detection",
            363       "id": "44adc011-570b-46cf-9a65-ab72ca690477",
            364       "name": "Object Detection",
            365       "state": "active",
            366       "vocabulary_id": null
            367     },
            368     {
            369       "display_name": "Predicate Estimation",
            370       "id": "42b00e50-cb78-48ff-b003-331f7e18c053",
            371       "name": "Predicate Estimation",
            372       "state": "active",
            373       "vocabulary_id": null
            374     },
            375     {
            376       "display_name": "Predicate Labeling",
            377       "id": "1eb3ff21-ffe4-4852-b334-5dbbaf84db43",
            378       "name": "Predicate Labeling",
            379       "state": "active",
            380       "vocabulary_id": null
            381     },
            382     {
            383       "display_name": "Region Descriptions",
            384       "id": "ab5f67c4-cedf-4402-bd40-b667f3b08141",
            385       "name": "Region Descriptions",
            386       "state": "active",
            387       "vocabulary_id": null
            388     },
            389     {
            390       "display_name": "Scene Graph",
            391       "id": "9a681371-0bc0-4158-a772-cb0a8bace01e",
            392       "name": "Scene Graph",
            393       "state": "active",
            394       "vocabulary_id": null
            395     },
            396     {
            397       "display_name": "Scene understanding",
            398       "id": "1da18641-2ea7-4cb0-9316-bf93df4fbba1",
            399       "name": "Scene understanding",
            400       "state": "active",
            401       "vocabulary_id": null
            402     },
            403     {
            404       "display_name": "Visual Attention",
            405       "id": "75d9febf-b2ed-45b2-8b35-90c67689b87b",
            406       "name": "Visual Attention",
            407       "state": "active",
            408       "vocabulary_id": null
            409     },
            410     {
            411       "display_name": "Visual Genome",
            412       "id": "e97cf598-66b5-4c61-9aba-b5b64456bc24",
            413       "name": "Visual Genome",
            414       "state": "active",
            415       "vocabulary_id": null
            416     },
            417     {
            418       "display_name": "Visual Grounding",
            419       "id": "82dad3b4-e1c2-47b9-9aa8-dc78c0274934",
            420       "name": "Visual Grounding",
            421       "state": "active",
            422       "vocabulary_id": null
            423     },
            424     {
            425       "display_name": "Visual Question Answering",
            426       "id": "d8aee55f-7e04-411f-8ce5-0794bc0406b0",
            427       "name": "Visual Question Answering",
            428       "state": "active",
            429       "vocabulary_id": null
            430     },
            431     {
            432       "display_name": "Visual Relationship Detection",
            433       "id": "f7edcdba-a41e-4e73-81d0-47895d9535f0",
            434       "name": "Visual Relationship Detection",
            435       "state": "active",
            436       "vocabulary_id": null
            437     },
            438     {
            439       "display_name": "Visual Relationship Prediction",
            440       "id": "d2c5b531-6b4f-426b-a7c7-ff6111807a74",
            441       "name": "Visual Relationship Prediction",
            442       "state": "active",
            443       "vocabulary_id": null
            444     },
            445     {
            446       "display_name": "Visual Semantic Graph",
            447       "id": "689c4bd2-664e-475c-910b-09e7842538cf",
            448       "name": "Visual Semantic Graph",
            449       "state": "active",
            450       "vocabulary_id": null
            451     },
            452     {
            453       "display_name": "Visual Understanding",
            454       "id": "29ed501b-787b-402e-85f8-86e1b4a4adde",
            455       "name": "Visual Understanding",
            456       "state": "active",
            457       "vocabulary_id": null
            458     },
            459     {
            460       "display_name": "attribute classification",
            461       "id": "2bb9ea17-d19a-4770-995a-d2b7b8273127",
            462       "name": "attribute classification",
            463       "state": "active",
            464       "vocabulary_id": null
            465     },
            466     {
            467       "display_name": "bounding box coordinates",
            468       "id": "f08c1824-7990-4879-b3af-9375508f1b4c",
            469       "name": "bounding box coordinates",
            470       "state": "active",
            471       "vocabulary_id": null
            472     },
            473     {
            474       "display_name": "conditional models",
            475       "id": "ccb12b57-29e2-448d-b130-db0dc2449577",
            476       "name": "conditional models",
            477       "state": "active",
            478       "vocabulary_id": null
            479     },
            480     {
            481       "display_name": "dense captioning",
            482       "id": "717a0f4f-fee3-487c-9303-5d38cc703f5d",
            483       "name": "dense captioning",
            484       "state": "active",
            485       "vocabulary_id": null
            486     },
            487     {
            488       "display_name": "description generation",
            489       "id": "cbb25ecc-f485-4021-a3bc-86814d479080",
            490       "name": "description generation",
            491       "state": "active",
            492       "vocabulary_id": null
            493     },
            494     {
            495       "display_name": "image annotation",
            496       "id": "6734bcb9-ad78-4289-8816-09b71832ab8e",
            497       "name": "image annotation",
            498       "state": "active",
            499       "vocabulary_id": null
            500     },
            501     {
            502       "display_name": "image captioning",
            503       "id": "f1bbe827-a03a-4280-b9fa-0599ccfc0541",
            504       "name": "image captioning",
            505       "state": "active",
            506       "vocabulary_id": null
            507     },
            508     {
            509       "display_name": "image description",
            510       "id": "99e3a7d7-7ee7-4f66-adf4-28ac19f60923",
            511       "name": "image description",
            512       "state": "active",
            513       "vocabulary_id": null
            514     },
            515     {
            516       "display_name": "image segmentation",
            517       "id": "7eaed78e-c73a-4929-a8c9-60265069f59a",
            518       "name": "image segmentation",
            519       "state": "active",
            520       "vocabulary_id": null
            521     },
            522     {
            523       "display_name": "image understanding",
            524       "id": "a209b790-086a-4345-b6af-b7b011b6e040",
            525       "name": "image understanding",
            526       "state": "active",
            527       "vocabulary_id": null
            528     },
            529     {
            530       "display_name": "image-text embedding",
            531       "id": "557b18b0-568a-4195-8359-58042c400d33",
            532       "name": "image-text embedding",
            533       "state": "active",
            534       "vocabulary_id": null
            535     },
            536     {
            537       "display_name": "image-text pairs",
            538       "id": "07cb3691-65c6-4b58-b3bf-5290980a86e3",
            539       "name": "image-text pairs",
            540       "state": "active",
            541       "vocabulary_id": null
            542     },
            543     {
            544       "display_name": "image-text retrieval",
            545       "id": "2ce62fa7-a31e-4be3-9312-92be69a8726b",
            546       "name": "image-text retrieval",
            547       "state": "active",
            548       "vocabulary_id": null
            549     },
            550     {
            551       "display_name": "keypoints",
            552       "id": "d8c2caaa-a318-4485-91a0-e9dc1826e52e",
            553       "name": "keypoints",
            554       "state": "active",
            555       "vocabulary_id": null
            556     },
            557     {
            558       "display_name": "labeling",
            559       "id": "5a92ec36-729e-4e7b-8719-df680a38d6d8",
            560       "name": "labeling",
            561       "state": "active",
            562       "vocabulary_id": null
            563     },
            564     {
            565       "display_name": "large-scale dataset",
            566       "id": "a9c694bf-f591-4625-a20e-d53d3f90d489",
            567       "name": "large-scale dataset",
            568       "state": "active",
            569       "vocabulary_id": null
            570     },
            571     {
            572       "display_name": "layout",
            573       "id": "0cf6a03b-7e94-49a7-b6a2-3d4204c308bb",
            574       "name": "layout",
            575       "state": "active",
            576       "vocabulary_id": null
            577     },
            578     {
            579       "display_name": "multiple entities",
            580       "id": "7810c6d2-47ba-43cd-93f2-74c41e5e35cf",
            581       "name": "multiple entities",
            582       "state": "active",
            583       "vocabulary_id": null
            584     },
            585     {
            586       "display_name": "object detection",
            587       "id": "607283c7-9e12-4167-9101-7f8078fb6537",
            588       "name": "object detection",
            589       "state": "active",
            590       "vocabulary_id": null
            591     },
            592     {
            593       "display_name": "pairwise relationships",
            594       "id": "cc61370a-a135-4b4c-8501-62313b4b8c22",
            595       "name": "pairwise relationships",
            596       "state": "active",
            597       "vocabulary_id": null
            598     },
            599     {
            600       "display_name": "person-in-context synthesis",
            601       "id": "ae487124-3926-4401-aa97-3bb78331ba42",
            602       "name": "person-in-context synthesis",
            603       "state": "active",
            604       "vocabulary_id": null
            605     },
            606     {
            607       "display_name": "phrase grounding",
            608       "id": "92b1c885-9229-4d53-8542-93d53a96e5c8",
            609       "name": "phrase grounding",
            610       "state": "active",
            611       "vocabulary_id": null
            612     },
            613     {
            614       "display_name": "question and answer pairs",
            615       "id": "adb1335d-1e4e-4630-8016-a22e6131ef71",
            616       "name": "question and answer pairs",
            617       "state": "active",
            618       "vocabulary_id": null
            619     },
            620     {
            621       "display_name": "question answering",
            622       "id": "d4e35d0b-ca07-4d05-bb27-50b2f30f2f00",
            623       "name": "question answering",
            624       "state": "active",
            625       "vocabulary_id": null
            626     },
            627     {
            628       "display_name": "scene graph",
            629       "id": "2687f43d-e44b-4991-8448-dc7791a3cfdf",
            630       "name": "scene graph",
            631       "state": "active",
            632       "vocabulary_id": null
            633     },
            634     {
            635       "display_name": "scene graph generation",
            636       "id": "2a0d95d7-11aa-4c4e-afbe-443469a9f78d",
            637       "name": "scene graph generation",
            638       "state": "active",
            639       "vocabulary_id": null
            640     },
            641     {
            642       "display_name": "scienti\ufb01c diagrams",
            643       "id": "331e1e79-382b-49a3-b864-a2ce03a6776a",
            644       "name": "scienti\ufb01c diagrams",
            645       "state": "active",
            646       "vocabulary_id": null
            647     },
            648     {
            649       "display_name": "semantic scene graphs",
            650       "id": "3fd97463-4078-413b-916a-8a074f8ee68a",
            651       "name": "semantic scene graphs",
            652       "state": "active",
            653       "vocabulary_id": null
            654     },
            655     {
            656       "display_name": "unpaired learning",
            657       "id": "2586ff50-5fc6-4dee-9b2b-792e507bfbf6",
            658       "name": "unpaired learning",
            659       "state": "active",
            660       "vocabulary_id": null
            661     },
            662     {
            663       "display_name": "vision and language",
            664       "id": "3745ee63-01b9-4847-8aeb-4d8a32d5b9f7",
            665       "name": "vision and language",
            666       "state": "active",
            667       "vocabulary_id": null
            668     },
            669     {
            670       "display_name": "vision-language pre-training",
            671       "id": "fd01e9f9-7fca-46ce-83d8-ff81c76d8e7a",
            672       "name": "vision-language pre-training",
            673       "state": "active",
            674       "vocabulary_id": null
            675     },
            676     {
            677       "display_name": "visual genome",
            678       "id": "81fb8015-731c-4366-9368-085564d34c07",
            679       "name": "visual genome",
            680       "state": "active",
            681       "vocabulary_id": null
            682     },
            683     {
            684       "display_name": "visual question answering",
            685       "id": "8be24887-e029-464f-b56c-1d0a460e2005",
            686       "name": "visual question answering",
            687       "state": "active",
            688       "vocabulary_id": null
            689     },
            690     {
            691       "display_name": "visual relation detection",
            692       "id": "59a14182-c50d-4c5f-8cff-40da79128e3d",
            693       "name": "visual relation detection",
            694       "state": "active",
            695       "vocabulary_id": null
            696     },
            697     {
            698       "display_name": "visual relationship detection",
            699       "id": "e1f83c99-b862-498b-b955-ee73b1648e6f",
            700       "name": "visual relationship detection",
            701       "state": "active",
            702       "vocabulary_id": null
            703     },
            704     {
            705       "display_name": "visual representations",
            706       "id": "26cfeecf-420e-429a-ba65-5d87d38f6048",
            707       "name": "visual representations",
            708       "state": "active",
            709       "vocabulary_id": null
            710     },
            711     {
            712       "display_name": "visual storytelling",
            713       "id": "e9c4b19d-67ce-43c9-ab9e-b6ca1068e53b",
            714       "name": "visual storytelling",
            715       "state": "active",
            716       "vocabulary_id": null
            717     },
            718     {
            719       "display_name": "visual-textual entailment",
            720       "id": "5140c115-a2da-45eb-8a74-88f8b7900790",
            721       "name": "visual-textual entailment",
            722       "state": "active",
            723       "vocabulary_id": null
            93     } 724     }
            94   ], 725   ],
            t 95   "services_used_list": "", t 
            96   "state": "active", 
            97   "tags": [], 
            98   "title": "Visual Genome", 726   "title": "Visual Genome",
            99   "type": "dataset", 727   "type": "dataset",
            100   "version": "" 728   "version": ""
            101 } 729 }