EVENTO
Understanding the Geometry of Image Databases Represented as Riemannian Manifolds Embedded into Image Spaces
Tipo de evento: Exame de Qualificação
Riemannian manifold concepts are important for deep learning models that can be trained to produce new data given a set of samples of an inherently geometric space named data manifold [1]. The main focus of this work is Riemannian manifold learning, where the training data is a sample of a Riemannian manifold. As a result, the geometry reconstruction can be carried out using operations that are guided by the datas presumptive geometric properties. Specifically, the data manifold is parameterized by differentiable maps that are computed through neural networks. Consequently, we can apply automatic differentiation techniques to calculate the derivatives necessary to study the geometric properties of [3,5]. Additionally, it has been shown that the latent space of a Generative Adversarial Network (GAN) can encode remarkable semantics in a few subspaces [4]. Thus, to be able to locate these related subspaces, researchers are used to explore statistical elements from a group of synthesized data, and these related subspaces are likely to regulate the properties of images on a global scale. In order to achieve more accurate control of GAN generation, in this work we aim to investigate these low-rank subspaces and study some of their relevant geometric properties. With the use of this geometric analysis of these subspaces, our goal is to achieve fine grain control over the precise editing of any images regions of interest. In this work, GANs are prominently taken into consideration as the foundation for deep learning models to learn a parametric representation of data manifolds in image spaces and subspaces that enable the recoveryof the geometric structure behind data samples [6].Para assistir acesse:meet.google.com/mjc-irzj-orx
Data Início: 16/12/2022 Hora: 09:00 Data Fim: 16/12/2022 Hora: 12:00
Local: LNCC - Laboratório Nacional de Computação Ciêntifica - Virtual
Aluno: Paulo Alves Braz - - LNCC
Orientador: Gilson Antônio Giraldi - Laboratório Nacional de Computação Científica - LNCC
Participante Banca Examinadora: Fabio Andre Machado Porto - Laboratório Nacional de Computação Científica - LNCC Raul Queiroz Feitosa - Pontifícia Universidade Católica do Rio de Janeiro - PUC-RIO Renato Portugal - Laboratório Nacional de Computação Científica - LNCC
Suplente Banca Examinadora: Antônio Tadeu Azevedo Gomes - Laboratório Nacional de Computação Científica - LNCC
