A Layer-Based Sequential Framework for Scene Generation with GANs

Mehmet Ozgur Turkoglu, Luuk Spreeuwers, William Thong, Berkay Kicanaoglu

    Research output: Contribution to conferencePaper

    2 Downloads (Pure)

    Abstract

    The visual world we sense, interpret and interact everyday is a complex composition of interleaved physical entities. Therefore, it is a very challenging task to generate vivid scenes of similar complexity using computers. In this work, we present a scene generation framework based on Generative Adversarial Networks (GANs) to sequentially compose a scene, breaking down the underlying problem into smaller ones. Different than the existing approaches, our framework offers an explicit control over the elements of a scene through separate background and foreground generators. Starting with an initially generated background, foreground objects then populate the scene one-by-one in a sequential manner. Via quantitative and qualitative experiments on a subset of the MS-COCO dataset, we show that our proposed framework produces not only more diverse images but also copes better with affine transformations and occlusion artifacts of foreground objects than its counterparts.
    Original languageEnglish
    Number of pages9
    Publication statusPublished - Feb 2019
    EventThirty-Third AAAI Conference on Artificial Intelligence (AAAI-19) - Hilton Hawaiian Village, Honolulu, United States
    Duration: 27 Jan 20191 Feb 2019
    https://aaai.org/Conferences/AAAI-19/

    Conference

    ConferenceThirty-Third AAAI Conference on Artificial Intelligence (AAAI-19)
    Abbreviated titleAAAI-19
    CountryUnited States
    CityHonolulu
    Period27/01/191/02/19
    Internet address

    Fingerprint

    Chemical analysis
    Experiments

    Cite this

    Turkoglu, M. O., Spreeuwers, L., Thong, W., & Kicanaoglu, B. (2019). A Layer-Based Sequential Framework for Scene Generation with GANs. Paper presented at Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19), Honolulu, United States.
    Turkoglu, Mehmet Ozgur ; Spreeuwers, Luuk ; Thong, William ; Kicanaoglu, Berkay. / A Layer-Based Sequential Framework for Scene Generation with GANs. Paper presented at Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19), Honolulu, United States.9 p.
    @conference{220643f09e2a43f598ddca25fac9c598,
    title = "A Layer-Based Sequential Framework for Scene Generation with GANs",
    abstract = "The visual world we sense, interpret and interact everyday is a complex composition of interleaved physical entities. Therefore, it is a very challenging task to generate vivid scenes of similar complexity using computers. In this work, we present a scene generation framework based on Generative Adversarial Networks (GANs) to sequentially compose a scene, breaking down the underlying problem into smaller ones. Different than the existing approaches, our framework offers an explicit control over the elements of a scene through separate background and foreground generators. Starting with an initially generated background, foreground objects then populate the scene one-by-one in a sequential manner. Via quantitative and qualitative experiments on a subset of the MS-COCO dataset, we show that our proposed framework produces not only more diverse images but also copes better with affine transformations and occlusion artifacts of foreground objects than its counterparts.",
    author = "Turkoglu, {Mehmet Ozgur} and Luuk Spreeuwers and William Thong and Berkay Kicanaoglu",
    year = "2019",
    month = "2",
    language = "English",
    note = "Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19), AAAI-19 ; Conference date: 27-01-2019 Through 01-02-2019",
    url = "https://aaai.org/Conferences/AAAI-19/",

    }

    Turkoglu, MO, Spreeuwers, L, Thong, W & Kicanaoglu, B 2019, 'A Layer-Based Sequential Framework for Scene Generation with GANs' Paper presented at Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19), Honolulu, United States, 27/01/19 - 1/02/19, .

    A Layer-Based Sequential Framework for Scene Generation with GANs. / Turkoglu, Mehmet Ozgur; Spreeuwers, Luuk; Thong, William; Kicanaoglu, Berkay.

    2019. Paper presented at Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19), Honolulu, United States.

    Research output: Contribution to conferencePaper

    TY - CONF

    T1 - A Layer-Based Sequential Framework for Scene Generation with GANs

    AU - Turkoglu, Mehmet Ozgur

    AU - Spreeuwers, Luuk

    AU - Thong, William

    AU - Kicanaoglu, Berkay

    PY - 2019/2

    Y1 - 2019/2

    N2 - The visual world we sense, interpret and interact everyday is a complex composition of interleaved physical entities. Therefore, it is a very challenging task to generate vivid scenes of similar complexity using computers. In this work, we present a scene generation framework based on Generative Adversarial Networks (GANs) to sequentially compose a scene, breaking down the underlying problem into smaller ones. Different than the existing approaches, our framework offers an explicit control over the elements of a scene through separate background and foreground generators. Starting with an initially generated background, foreground objects then populate the scene one-by-one in a sequential manner. Via quantitative and qualitative experiments on a subset of the MS-COCO dataset, we show that our proposed framework produces not only more diverse images but also copes better with affine transformations and occlusion artifacts of foreground objects than its counterparts.

    AB - The visual world we sense, interpret and interact everyday is a complex composition of interleaved physical entities. Therefore, it is a very challenging task to generate vivid scenes of similar complexity using computers. In this work, we present a scene generation framework based on Generative Adversarial Networks (GANs) to sequentially compose a scene, breaking down the underlying problem into smaller ones. Different than the existing approaches, our framework offers an explicit control over the elements of a scene through separate background and foreground generators. Starting with an initially generated background, foreground objects then populate the scene one-by-one in a sequential manner. Via quantitative and qualitative experiments on a subset of the MS-COCO dataset, we show that our proposed framework produces not only more diverse images but also copes better with affine transformations and occlusion artifacts of foreground objects than its counterparts.

    M3 - Paper

    ER -

    Turkoglu MO, Spreeuwers L, Thong W, Kicanaoglu B. A Layer-Based Sequential Framework for Scene Generation with GANs. 2019. Paper presented at Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19), Honolulu, United States.