Yliess Hati
Generalist Engineer, Ph.D. Student in Deep Learning, Electronics Enthousiast
I like to create interactive experiences powered by deep neural nets!
Professional Experience
Ph.D. Student
De Vinci Innovation Center (DVIC)
Paris La Défense, France
- Research on Deep Learning powered Creative Tools such as user-guided Automatic Colorization, audio-driven Facial Animation and Reconstruction, and audio-driven Dream Exploration.
- Teaching modern Artificial Intelligence (AI), Deep Learning (DL), and Computer Graphics (CG).
- Coaching Creative Technologist Students at the DVIC.
- Ensured student demos Quality Control.
- Implemented and released a real-time Style Transfer integrated camera frame called MagicWall, one of the first DVIC Consulting Contract.
Final Year Internship
De Vinci Innovation Center (DVIC)
Paris La Défense, France
- Implemented, wrote, and submitted my second publication on a User-Guided Anime Linear Colorization Tool to the Conference on Visual Media Production (CVMP).
- Helped design and prepare the DVIC first official curriculum and infrastructure deployment.
Visiting Student
MIT Medial Lab - Responsive Environments Group
Boston, MA, USA
- Implemented a web app for a crowdsourced panda pose video labelization.
- Implemented and trained a pandas pose estimator for state classification in collaboration with the San Diego Zoo.
- Conducted a user-study, wrote, and submitted my first publication on an Audio-Driven 3D Mouth Animation Interface for Personal Assistants for the International Conference on Auditory Display (ICAD).
- Helped implement a contrastive learning scheme using triplet loss for a wildlife audio classifier Tid'Zam deployed in Tidmarsh.
Internship
Adaltas
Suresnes, France
- Implemented a State-of-the-Art Text Summarization Pipeline based on Pointer Generator with Attention we named Lando.
- Implemented a Full Stack pipeline to deploy the Summarization model on a Web Application using Continuous Integration and Deployment using GitLab, Docker, and Kubernetes.
- Wrote conference reports and articles on the company website.
Technical Skills
Programming
I started programming in middle school as a hobby and have not stopped since. The language is not the barrier, I can adapt and learn. I am familiar with the inner working of computers, operating systems, compilers, interpreters, and such.
Deep Learning
Since 2015, I have been learning and playing with Neural Nets. I am excited by the high pace and the open-source mentality of the Deep Learning community. To me, Deep Learning is a powerful tool to have and master. This technology has shown remarkable promises in many fields and has yet to show its full potential.
Computer Graphics
I love Video Games, VFX, and digital art. Computer Graphics is also responsible for the GPUs I love and use in my daily work. For these reasons, I have invested quite some time in this topic. I used to program Game/Rendering Engines as a hobby.
Full Stack Development
I have started my programming journey by building websites for fun. I am familiar with modern web development frameworks. Web is an easy to deploy platform for building and sharing experiences. I have some experience in designing full stack applications.
DevOps
I use Linux and Git as my daily driver. I have some experience with Docker and Kuernetes although not as proficient with the latter. This website is hosted on GitHub Pages and is built using CI/CD.
Digital Media Art
I grew up loving drawing and digital art. I am proficient with both 2D and 3D art techniques.
Electronics
Being a Generalist Engineer, I have knowledge in electronics First Principles. I have recently started playing with building, training, and integrating Tiny Machine Learning applications into Micro Controllers. Hope to do more on this topic
Languages
French is my native language. I use English at work and in my daily media consumption.
Teaching
Instructor
Computer Graphics (CG)
De Vinci Innovation Center (DVIC)
Paris La Défense, France
Teaching the basics of Computer Graphics and Video Game Programming to Creative Technologist Students.
Instructor
Artificial Intelligence (AI)
De Vinci Innovation Center (DVIC)
Paris La Défense, France
Teaching Deep Learning from First Principles to SOTA Applications to Creative Technologist Students.
Instructor
Artificial Intelligence (AI) for the Internet of Things (IOT)
Ecole d'Ingenieur Leonard De Vinci (ESILV)
Paris La Défense, France
Teaching the basics of Deep Learning and how to deploy models for Embedded Devices and IOT Applications.
Teaching Assistant
Machine Learning (ML) in R
Ecole d'Ingenieur Leonard De Vinci (ESILV)
Paris La Défense, France
Teaching Machine Learning, Pattern Recognition, and Data Science in R to Engineering Students.
Teaching Assistant
Design Patterns in C# (.NET)
Ecole d'Ingenieur Leonard De Vinci (ESILV)
Paris La Défense, France
Teaching Object-Oriented Programming (OOP) and Design Pattern in C# (.NET) to Engineering Students.
Teaching Assistant
Master Class on Artificial Intelligence (AI)
Ecole d'Ingenieur Leonard De Vinci (ESILV)
Paris La Défense, France
Learning and teaching SOTA Artificial Intelligence (AI), Deep Learning (DL) to Engineering Students.
Academic Publications
StencilTorch: an Iterative and User-Guided Framework for Anime Lineart Colorization
Image and Vision Computing New Zealand (IVCNZ) - 2022
Yliess HatiVincent TheveninFlorent NolotFrancis RousseauxClement Duhart
coming soon
This paper presents StencilTorch, a new approach to the task of automatic lineart colorization in computer vision. It is an interactive and user-guided framework that generates illustrations from a given lineart, color hints, and a mask, allowing for iterative workflows where the output of the first pass becomes the input of a second. This approach improves upon previous methods in both objective and subjective evaluations.
PaintsTorch: a User-Guided Anime Line Art Colorization Tool with Double Generator Conditional Adversarial Network
European Conference on Visual Media Production (CVMP) - 2019
Yliess HatiGregor JouetFrancis RousseauxClement Duhart
10.1145/3359998.3369401
This paper presents a new approach for user guided-colorization of line arts in computer vision. The current state of the art uses GANs to generate realistic and precise colorization, but relies on randomly sampled pixels as color hints for training. This approach aims to improve upon this by introducing a stroke simulation based approach for hint generation, making the model more robust to messy inputs. We also propose a new cleaner dataset, and explore the use of a double generator GAN to improve visual fidelity.
Text-Driven Mouth Animation for Human Computer Interaction with Personal Assistant
International Conference on Auditory Display (ICAD) - 2019
Yliess HatiFrancis RousseauxClement Duhart
10.21785/icad2019.032
This paper proposes a new neural architecture for text-driven 3D mouth animations for personal assistants to improve the realism and naturalness of the interactions. The study shows that such visual feedback improves users comfort for 78% of the candidates significantly while slightly improving their time perception.
Academic Awards
PI2 Project
Ecole d'Ingenieur Leonard De Vinci (ESILV)
Second Place
ADMA: Audio‑Driven Mouth Animations. Poster presentation and Demo stand.
PING Project
Ecole d'Ingenieur Leonard De Vinci (ESILV)
First Place
Project of the Year. Sun Wu Kong: Audio Queries for Video Game Interaction Poster presentation and Demo stand.
Education
Ph.D. Student
URCA CreSTIC - De Vinci Innovation Center (DVIC)
Reims - Paris La Défense, France
Ph.D. in Artificial Intelligence. Working on Deep Learning Assisted Creative-Expression applied to User-Guided Automatic Lineart Colorization.
Innovation Track
De Vinci Innovation Center (DVIC)
Paris La Défense, France
Experimental Innovation Track from the DVIC following the "Learning by Doing" and "Demo or Die" principals from the MIT MediaLab.
Generalist Engineering Degree
Ecole Superieure d'Ingenieur Leonard De Vinci (ESILV)
Paris La Défense, France
Majored in Computer Science, Big Data, and Internet of Thins (IBO).