Datasets for graphs

last modified : 19-07-2019

Here are describe the dataset that can be used when graphs are used as input (graph to image, image to graph, ...)

Visual Genome

  • Contains:

    • Objects: Lots of stuff, check the website
    • Relations: Lots too, check the website too
    • Regions: (such as sky, grass, ...) same as the two above

  • Size:

    • Total images: 108,077
    • Average image size (px): 500.14
    • Maximum image size (px): 1,280
    • Minimum image size (px): 72
    • Total region descriptions: 4,297,502
    • Total image object instances: 1,366,673
    • Unique image objects: 75,729
    • Total object-object relationship instances: 1,531,448
    • Unique relationships: 40,480
    • Total attribute-object instances: 1,670,182
    • Unique attributes: 40,513
    • Total Scene Graphs: 108,249
    • Total Region Graphs: 3,788,715

  • Article:

    • Title: Visual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotations
    • Authors: Ranjay Krishna, Yuke ZhuOliver, GrothJustin, JohnsonKenji, HataJoshua, KravitzStephanie, ChenYannis, Kalantidis, Li-Jia Li, David A. Shamma, Michael S. Bernstein, Li Fei-Fei
    • Link: article

  • Dataset: here

COCO-Stuff

  • Contains:

    • Dense pixel-level annotations
    • 80 thing classes, 91 stuff classes and 1 class 'unlabeled'
    • Instance-level annotations for things from COCO
    • Complex spatial context between stuff and things
    • 5 captions per image from COCO

  • Size:

    • 164K complex images from COCO

  • Article:

    • Title: COCO-Stuff: Thing and Stuff Classes in Context
    • Authors: H. Caesar, J. Uijlings, V. Ferrari
    • Link: article

  • Dataset: here