Cristhian Forigua
A5057972851- Human Pose and Action Recognition
- Computational Physics and Python Applications
- Advanced Vision and Imaging
- Anomaly Detection Techniques and Applications
- Multimodal Machine Learning Applications
- Single-cell and spatial transcriptomics
- Advanced Neural Network Applications
- Human-Automation Interaction and Safety
- Gene expression and cancer classification
- Virtual Reality Applications and Impacts
- Robotics and Sensor-Based Localization
- Cell Image Analysis Techniques
- Advanced Image Processing Techniques
Universidad de Los Andes
2024
Universidad de Los Andes
2022
The identification of cell types is a basic step the pipeline for Single-Cell RNA sequencing data analysis. However, unsupervised clustering cells from scRNA-seq has multiple challenges: high dimensional nature data, sparse gene expression matrix, and presence technical noise that can introduce false zero entries. In this study, we new algorithms data. first algorithm builds k-MST graph distances obtained directly input without dimensionality reduction. computation follows an iterative...
Accurately estimating and forecasting human body pose is important for enhancing the user's sense of immersion in Augmented Reality. Addressing this need, our paper introduces EgoCast, a bimodal method 3D using egocentric videos proprioceptive data. We study task realistic setting, extending boundaries temporal dynamic scenes building on current framework estimation wild. introduce current-frame module that generates pseudo-groundtruth poses inference, eliminating need past groundtruth...
We present EgoCOL, an egocentric camera pose estimation method for open-world 3D object localization. Our leverages sparse reconstructions in a two-fold manner, video and scan independently, to estimate the of frames renders with high recall precision. extensively evaluate our on Visual Query (VQ) localization Ego4D benchmark. EgoCOL can 62% 59% more poses than baseline Queries Localization challenge at CVPR 2023 val test sets, respectively. code is publicly available...
We present Ego-Exo4D, a diverse, large-scale multimodal multiview video dataset and benchmark challenge. Ego-Exo4D centers around simultaneously-captured egocentric exocentric of skilled human activities (e.g., sports, music, dance, bike repair). More than 800 participants from 13 cities worldwide performed these in 131 different natural scene contexts, yielding long-form captures 1 to 42 minutes each 1,422 hours combined. The nature the is unprecedented: accompanied by multichannel audio,...