Name:
Generative Models for Neural Fields.pdf
Size:
64.78Mb
Format:
PDF
Description:
Generative Models for Neural Fields - PhD thesis
Type
DissertationAuthors
Skorokhodov, Ivan
Advisors
Wonka, Peter
Elhoseiny, Mohamed

Committee members
Ghanem, Bernard
Heidrich, Wolfgang

Niessner, Matthias
Program
Computer ScienceDate
2023-02Permanent link to this record
http://hdl.handle.net/10754/690419
Metadata
Show full item recordAbstract
Deep generative models are deep learning-based methods that are optimized to synthesize samples of a given distribution. During the past years, they have attracted a lot of interest from the research community, and the developed tools now enjoy many practical applications in content creation and editing. In computer vision, such models are typically built for images, videos, and 3D objects. Recently, there has emerged a paradigm of neural fields, which unifies the representations of such types of data by parametrizing them via neural networks. In this work, we develop generative modeling methods for images, videos, and 3D objects which treat the underlying data in such a form. We show that this perspective can yield state-of-the-art synthesis quality and many useful practical benefits, like interpolation/extrapolation capabilities, geometric inductive biases, and more efficient training and inference.Citation
Skorokhodov, I. (2023). Generative Models for Neural Fields [KAUST Research Repository]. https://doi.org/10.25781/KAUST-54DV8ae974a485f413a2113503eed53cd6c53
10.25781/KAUST-54DV8
Scopus Count
Related items
Showing items related by title, author, creator and subject.
-
Reduced resilience of a globally distributed coccolithophore to ocean acidification: Confirmed up to 2000 generations, supplement to: Jin, Peng; Gao, Kunshan (2016): Reduced resilience of a globally distributed coccolithophore to ocean acidification: Confirmed up to 2000 generations. Marine Pollution Bulletin, 103(1-2), 101-108Jin, Peng; Gao, Kunshan (PANGAEA - Data Publisher for Earth & Environmental Science, 2016) [Dataset]Ocean acidification (OA), induced by rapid anthropogenic CO2 rise and its dissolution in seawater, is known to have consequences for marine organisms. However, knowledge on the evolutionary responses of phytoplankton to OA has been poorly studied. Here we examined the coccolithophore Gephyrocapsa oceanica, while growing it for 2000 generations under ambient and elevated CO2 levels. While OA stimulated growth in the earlier selection period (from generations 700 to 1550), it reduced it in the later selection period up to 2000 generations. Similarly, stimulated production of particulate organic carbon and nitrogen reduced with increasing selection period and decreased under OA up to 2000 generations. The specific adaptation of growth to OA disappeared in generations 1700 to 2000 when compared with that at 1000 generations. Both phenotypic plasticity and fitness decreased within selection time, suggesting that the species' resilience to OA decreased after 2000 generations under high CO2 selection.
-
M-Elfeki/GDPP: Generator loss to reduce mode-collapse and to improve the generated samples quality.Elfeki, Mohamed; Couprie, Camille; Elhoseiny, Mohamed; Rivière, Morgane (Github, 2018-11-30) [Software]Generator loss to reduce mode-collapse and to improve the generated samples quality.
-
Entropy generation minimization: A practical approach for performance evaluation of temperature cascaded co-generation plantsMyat, Aung; Thu, Kyaw; Kim, Youngdeuk; Saha, Bidyut Baran; Ng, K. C. (Energy, Elsevier BV, 2012-10) [Article]We present a practical tool that employs entropy generation minimization (EGM) approach for an in-depth performance evaluation of a co-generation plant with a temperature-cascaded concept. Co-generation plant produces useful effect production sequentially, i.e., (i) electricity from the micro-turbines, (ii) low pressure steam at 250 °C or about 8-10 bars, (iii) cooling capacity of 4 refrigeration tones (Rtons) and (iv) dehumidification of outdoor air for air conditioned space. The main objective is to configure the most efficient configuration of producing power and heat. We employed entropy generation minimization (EGM) which reflects to minimize the dissipative losses and maximize the cycle efficiency of the individual thermally activated systems. The minimization of dissipative losses or EGM is performed in two steps namely, (i) adjusting heat source temperatures for the heat-fired cycles and (ii) the use of Genetic Algorithm (GA), to seek out the sensitivity of heat transfer areas, flow rates of working fluids, inlet temperatures of heat sources and coolant, etc., over the anticipated range of operation to achieve maximum efficiency. With EGM equipped with GA, we verified that the local minimization of entropy generation individually at each of the heat-activated processes would lead to the maximum efficiency of the system. © 2012.