Abdallah Dib - Research Scientist

Recent News

06/07/2025

We've published a new blog post on Ubisoft website, showcasing our work on the paper MoSAR, a technique used by artists to streamline their workflow.

25/06/2025

SEREP paper got accepted to ICCV 2025. We introduce a novel learning-based method for monocular facial expression capture and retargeting. More details from Here

20/05/2025

One paper accepted to 'AI for Creative Visual Content Generation Editing and Understanding' (CVEU), CVPR 2025. We propose a texture generator model giving artists control over shape, skin tone and fine details. More details from Here

09/03/2024

We released FFHQ-UV-Intrinsics dataset that contains intrinsics texture maps for 10K subjects at HD resolution. Download it from here

01/03/2024

Mosar paper got accepted to CVPR 2024. MoSAR turns a portrait image into a relightable 3D avatar. More details from Here

05/04/2023

We published a technical paper showcasing our FaceLab solution, which was used by artists to capture 3D facial performance for the 2019 film Cats.

01/02/2023

S2F2 paper got accepted to FG2023. S2F2 is a robust self-supervised model that estimate 3D shape and reflectance from a monocular image. More details from Here

18/05/2022

DeepNextFace is a 3D face reconstruction library from a single monocular RGB image via deep convolutional neural networks and differentiable ray tracing. Check it from https://github.com/abdallahdib/DeepNextFace

21/04/2022

NextFace is a lightweight open source library, written in pytorch, for high fidelity face reconstruction. Check it from https://github.com/abdallahdib/NextFace

11/10/2021

Our paper on self-supervised monocular 3D face reconstruction got accepted to ICCV 2021. More details from Here

09/03/2021

Our paper on monocular 3D face reconstruction got accepted to EuroGraphics 2021. We achieve realistic 3D face reconstruction from a single image. More details from Here

I am a research scientist working at Ubisoft La Forge (Montreal). Previously, I worked as a Senior Research Scientist at InterDigital R&I (France) and Technicolor Research, and as a Research Engineer at Inria. I obtained my PhD from Inria in 2016.

I develop methods for face capture, facial animation, and avatar reconstruction for video games and interactive entertainment. My passion lies in conducting applied research that tackles real-world production challenges, identifying workflow bottlenecks to developing novel techniques that become integrated tools, enabling more efficient workflows for artists and creators.

Research Interests: Computer Vision, Computer Graphics, Machine Learning, Generative Models, Neural Rendering, Self-Supervised Learning, Disentangled representations

Contact: deeb.abdallah [at] gmail [dot] com

I supervise research interns and PhD students, and I am always seeking motivated candidates to contribute to state-of-the-art projects.

Selected Projects

A blog post published on Ubisoft website, showcasing our work on MoSAR, a novel technique that transforms a single portrait photo into a fully relightable and detailed 3D avatar, streamlining character creation workflows for artists.

A technical paper presents FaceLab, our facial performance capture solution that enabled MPC artists to automatically extract facial performances from single camera footage. This technology successfully processed over 260 shots for the 2019 film Cats.

NextFace, my open-source PyTorch library for high-fidelity 3D face reconstruction from monocular RGB image. NextFace estimates facial attributes including 3D geometry, PBR materials (diffuse, specular, roughness), camera parameters, and illumination. With over 770 stars and 100 forks, NextFace serves as a state-of-the-art baseline for the computer vision research community.

We release FFHQ-UV-Intrinsics, a unique dataset containing intrinsic maps for 10,000 subjects including: diffuse, specular, ambient occlusion, translucency, and normal maps. The dataset is Available publicly for research use from here.

Selected Publications

SEREP: Semantic Facial Expression Representation for Robust In-the-Wild Capture and Retargeting

Arthur Josi, Luiz Gustavo Hafemann, Abdallah Dib, Emeline Got, Rafael MO Cruz, Marc-Andre Carbonneau

International Conference on Computer Vision (ICCV 2025)

Paper Project

Geometry-Aware Texture Generation for 3D Head Modeling with Artist-driven Control

Amin Fadaeinejad, Abdallah Dib, Luiz Gustavo Hafemann, Emeline Got, Trevor Anderson, Amaury Depierre, Nikolaus F. Troje, Marcus A Brubaker, Marc-Andre Carbonneau

AI for Creative Visual Content Generation Editing and Understanding (CVEU), CVPR 2025

Paper Project

MoSAR: Monocular Semi-Supervised Model for Avatar Reconstruction using Differentiable Shading

Abdallah Dib, Luiz Gustavo Hafemann, Emeline Got, Trevor Anderson, Amin Fadaeinejad, Rafael M. O. Cruz, Marc-Andre Carbonneau

IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR 2024)

Paper Project Dataset

S2F2: Self-Supervised High Fidelity Face Reconstruction From Monocular Image

Abdallah Dib, Cedric Thebault, Junghyun Ahn, Philippe-Henri Gosselin, Louis Chevallier

International Conference on Automatic Face and Gesture Recognition FG 2023

Paper Video

Practical Face Reconstruction via Differentiable Ray Tracing

Abdallah Dib, Gaurav Bharaj, Junghyun Ahn, Cédric Thébault, Philippe-Henri Gosselin, Marco Romeo, Louis Chevallier

Computer Graphics Forum, Eurographics' 2021

Paper Video Code

Towards High Fidelity Monocular Face Reconstruction with Rich Reflectance using Self-supervised Learning and Ray Tracing

Abdallah Dib, Cedric Thebault, Junghyun Ahn, Philippe-Henri Gosselin, Christian Theobalt, Louis Chevallier

International Conference on Computer Vision (ICCV 2021)

Paper Video Code

PhotoApp: Photorealistic Appearance Editing of Head Portraits

Mallikarjun B R, Ayush Tewari, Abdallah Dib, Tim Weyrich, Bernd Bickel, Hans-Peter Seidel, Hanspeter Pfister, Wojciech Matusik, Louis Chevallier, Mohamed Elgharib, Christian Theobalt

ACM Transactions on Graphics (SIGGRAPH Asia 2021)

Paper Project