Jeffri Murrugarra-Llerena

I am a PhD student in the Computer Science department at Stony Brook University, advised by Prof. Paola Cascante-Bonilla and part of the SPEL lab. My research is focused on Computer vision and Natural Language Processing. Before this, I was as an assistant researcher at Federal University of Rio Grande do Sul.

In 2022, I obtained my Masters in Computer Science from Federal University of Rio Grande do Sul, where I worked with Prof. Claudio Rosito Jung at the department of informatics. Our research lies problems about spherical images and oriented object detectors. Previously, I obtained a Bachelor of Science in Computer Science from National University of Trujillo. During my undergrad studies I did an exchange bachelor for a year at the University of Sao Paulo (ICMC).

Email  /  CV  /  Linkedin  /  Google Scholar  /  Github

• ...
• [Feb 2025] Paper accepted at CVPR 2025!!.
• [Oct 2024] Paper accepted at WACV 2025!!.
• [Aug 2024] Started my PhD at Stony Brook University.

GauCho detects oriented objects with typical representations or Oriented Ellipses. It relies on a novel regression head designed to directly predict the parameters of 2D Gaussian distributions through the Cholesky decomposition of their covariance matrices, which theoretically mitigates the boundary discontinuity problem.

The main goals of this work are to quantify the extent of annotation noise in terms of corner-wise discrepancies, assess how it impacts evaluation met-rics for object detection, and propose noise-aware alternatives that serve as upper and lower bounds.

We propose a new probabilistic loss function, to train and evaluate oriented object detectors named ProbIoU. ProbIoU have desired properties such as: scale invariant, metric properties, few hyperparameters and easy implementations.

We have implemented an attention module in conjunction with metric learning to enhance human interpretability, mirroring core courses per computing career.

This paper presented a critical analysis of popular datasets for HBB and OBB object detection, namely COCO, VOC, and DOTA/iSAID, aiming to check the consistency of bounding box annotations and segmentation masks and how discrepancies affect the IoU and AP metrics.

In this paper, we presented a comparative analysis of seven keypoint matching algorithms applied to 360◦ image pairs using several pose estimation approaches.

This paper provides a comprehensive survey on pioneer and state-of-the-art 3D scene geometry estimation methodologies based on single, two, or multiple images captured under the omnidirectional optics.

Credits