Source author record

Mijeong Kim

Mijeong Kim appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Computer Vision eess.IV Graphics Methodology

Catalog footprint

What is connected

2works

4topics

2close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2023arXiv

Appropriate use of parametric and nonparametric methods in estimating regression models with various shapes of errors

In this paper, a practical estimation method for a regression model is proposed using semiparametric efficient score functions applicable to data with various shapes of errors. First, I derive semiparametric efficient score vectors for a homoscedastic regression model without any assumptions of errors. Next, the semiparametric efficient score function can be modified assuming a certain parametric distribution of errors according to the shape of the error distribution or by estimating the error distribution non-parametrically. Nonparametric methods for errors can be used to estimate the parameters of interest or to find an appropriate parametric error distribution. In this regard, the proposed estimation methods utilize both parametric and nonparametric methods for errors appropriately. Through numerical studies, the performance of the proposed estimation methods is demonstrated.

preprint2022arXiv

InfoNeRF: Ray Entropy Minimization for Few-Shot Neural Volume Rendering

We present an information-theoretic regularization technique for few-shot novel view synthesis based on neural implicit representation. The proposed approach minimizes potential reconstruction inconsistency that happens due to insufficient viewpoints by imposing the entropy constraint of the density in each ray. In addition, to alleviate the potential degenerate issue when all training images are acquired from almost redundant viewpoints, we further incorporate the spatially smoothness constraint into the estimated images by restricting information gains from a pair of rays with slightly different viewpoints. The main idea of our algorithm is to make reconstructed scenes compact along individual rays and consistent across rays in the neighborhood. The proposed regularizers can be plugged into most of existing neural volume rendering techniques based on NeRF in a straightforward way. Despite its simplicity, we achieve consistently improved performance compared to existing neural view synthesis methods by large margins on multiple standard benchmarks.