Unai Elordi

I'm

About

AI and Computer Vision researcher at Vicomtech on Inteligent Transport System. Active science communicator for media sources such as radio and online events (euskal encounter).

AI and Computer Vision Researcher at Vicomtech

  • Website: www.vicomtech.org
  • City: Donostia/San Sebastian
  • Degree: Computer science degree, Master, PHD candidate
  • email: uelordi@vicomtech.org
  • ORCID:0000-0002-6276-01230

He loves computer vision and Deep Learning optimization for embedded devices such as (raspberry-pis jetson devices, google coral, intel VPUs). His greatest virtue is the versatility to learn new languages and software architectures. Member of the Google Associate Android Developer. Active science communicator.

Facts

Projects

Research papers

Science communications for media

Skills

Deep Learning Inference Optimization 100%
OpenCV 80%
TensorRT 60%
OpenVINO 80%
Serverless computing 60%
JavaScript 75%
Python 75%
C++ 80%
PHP 50%

Research

  • Unai Elordi, Alvaro Bertelsen, Luis Unzueta, Nerea Aranjuelo, Jon Goenetxea, Ignacio Arganda-Carreras (2021-Best Sudent Paper at KES2021). Optimal deployment of face recognition solutions in a heterogeneous IoT platform for secure elderly care applications. Procedia Computer Science., 192(1), 3204-3213.
  • Unai Elordi and Luis Unzueta and Jon Goenetxea and Sergio Sanchez-Carballido and Ignacio Arganda-Carreras and Oihana Otaegui (2021). Benchmarking Deep Neural Network Inference Performance on Serverless Environments With MLPerf. IEEE Softw., 38(1), 81–87.
  • Unai Elordi and Luis Unzueta and Jon Goenetxea and Estibaliz Loyo and Ignacio Arganda-Carreras and Oihana Otaegui (2021). On-demand Serverless Video Surveillance with Optimal Deployment of Deep Neural Networks. In Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2021, Volume 4: VISAPP, Online Streaming, February 8-10, 2021 (pp. 717–723).
  • Nerea Aranjuelo and Jorge Garcia and Luis Unzueta and Sara Garcia and Unai Elordi and Oihana Otaegui (2021). Building Synthetic Simulated Environments for Configuring and Training Multi-camera Systems for Surveillance Applications. In Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2021, Volume 5: VISAPP, Online Streaming, February 8-10, 2021 (pp. 80–91)
  • Jon Goenetxea and Luis Unzueta and Unai Elordi and Oihana Otaegui and Fadi Dornaika (2021). Efficient Multi-task based Facial Landmark and Gesture Detection in Monocular Images. In Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, VISIGRAPP 2021, Volume 5: VISAPP, Online Streaming, February 8-10, 2021 (pp. 680–687).
  • Luis Unzueta and Sandra Garcia and Jorge Garcia and Valentin Corbin and Nerea Aranjuelo and Unai Elordi and Oihana Otaegui and Maxime Danielli (2020). Building a Camera-based Smart Sensing System for Digitalized On-demand Aircraft Cabin Readiness Verification. In Proceedings of the International Conference on Robotics, Computer Vision and Intelligent Systems, ROBOVIS 2020, Budapest, Hungary, November 4-6, 2020 (pp. 98–105).
  • Goenetxea, J., Unzueta, L., Elordi, U., Ortega, J. D., & Otaegui, O. (2018). Efficient monocular point-of-gaze estimation on multiple screens and 3D face tracking for driver behaviour analysis . In 6th Int. Conf. on Driver Distraction and Inattention (pp. 1-8).
  • Unai Elordi and Luis Unzueta and Ignacio Arganda-Carreras and Oihana Otaegui (2018). How Can Deep Neural Networks Be Generated Efficiently for Devices with Limited Resources?. In Articulated Motion and Deformable Objects - 10th International Conference, AMDO 2018, Palma de Mallorca, Spain, July 12-13, 2018, Proceedings (pp. 24–33).
  • Clemotte, A., Arregui, H., Velasco, M. A., Unzueta, L., Goenetxea, J., Elordi, U., ... & Jauregui, E. TRAJECTORY CLUSTERING FOR THE CLASSIFICATION OF EYE-TRACKING USERS WITH MOTOR DISORDERS.
  • Leonardo De-Maeztu and Unai Elordi and Marcos Nieto and Javier Barandiarán and Oihana Otaegui (2015). A temporally consistent grid-based visual odometry framework for multi-core architectures . J. Real Time Image Process., 10(4), 759–769.
  • Elordi, U., Segura, A., Goenetxea, J., Moreno, A., & Arambarri, J. (2012, July). Virtual reality interfaces applied to web-based 3D E-commerce. In Engineering Systems Design and Analysis (Vol. 44847, pp. 341-350). American Society of Mechanical Engineers.

Resume

Sumary

Researcher

Researcher and developer with more than 11 years of experience in Computer Vision and Artificial Ingelligence algorithms. His research lines are Deep Learning inference optimization, Face gesture/recognition/verification, DeepFake detection, Video surveillance aplications.

  • Parque Científico y Tecnológico de Gipuzkoa, Paseo Mikeletegi 57,
  • 20009 Donostia / San Sebastián (España)
  • uelordi@vicomtech.org

Education

Phd candidate in Artificial Intelligent

2018-Current

EHU/UPV university

Universidad del País Vasco / Euskal Herriko Unibertsitatea. Barrio Sarriena s/n.48940 Leioa Bizkaia

Android associated developer

2017

Udacity

Master of Intelligent Systems and Computation Engineering

2015 - 2016

EHU/UPV university

Universidad del País Vasco / Euskal Herriko Unibertsitatea. Barrio Sarriena s/n.48940 Leioa Bizkaia

Computer Science Engineer

2003-2009

Mondragon Unibertsitatea (MGEP), Mondragon, Spain

Professional Experience

Researcher

2009 - Present

Vicomtech (Intelligent transport Systems (ITS))

  • Research about Deep Learnng optimizations to leverage optimal deployment in low-resource and embedded devices
  • Project Manager in public funded research projects
  • Active colaborator in open source projects (github)
  • Computer Vision and AI development in several areas (Face analysis, Face recognition, object detection, eye gaze estimation )

Science comunicator

Since 2019, Unai has been contributing to the traditional the current media sources to spread the science algorithms for the general public.

  • Euskadi irratia
  • Radio euskadi
  • Onda Vasca
  • Euskal Telebista
  • Cadena SER
  • Naiz irratia
  • twitch
  • berria
  • Euskal Encounter

Cofounder

2012-2013

Web Gluup S.L

  • Agile Teams leading
  • Email deliverability control
  • proactive web developer
  • Digital marketing enthusiast(landing strategies, search engine algorithm trackers)

Cofounder

2012-2013

Web Gluup S.L

  • Agile Teams leading
  • Email deliverability control
  • proactive web developer
  • Digital marketing enthusiast(landing strategies, search engine algorithm trackers)

Most relevant Open Source contributions

Face swap

Faceswap algorithm with texture triangulation approach. Dlib and MediaPipe 468 facemesh models are available.

MLPerf for serverless

This repo is the implementation of the Benchmarking DNN inference performance on serverless environment with MLPerf paper deployed to AWS Lambda.

Paper

Call_RenderScript_from_NDK

A simple example to enable RenderScript with gradle + cmake+ ndk

OpenCV 3.4 android

This repository is created to avoid the gradle problems with the OpenCV java sdk, usually if you are going to use OpenCV CameraBridge.

Basicpolycreator

This project is based on canvas HTML5 basic poligon drawing.

Contact

Location:

Vicomtech, Parque Científico y Tecnológico de Gipuzkoa, Paseo Mikeletegi 57, 20009 Donostia / San Sebastián (España)

Call:

+(34) 943 309 230