CV
Basics
| Name | Alexandre Symeonidis-Herzig |
| Label | PhD Researcher in Computer Vision |
| a.symeonidisherzig@gmail.com | |
| Url | https://github.com/ash-fortyfive |
| Summary | PhD researcher in computer vision, specializing in sign language production, photorealistic avatar synthesis, and human motion generation. Experienced in developing end-to-end deep learning systems, from multi-view data capture and preprocessing to model training and evaluation. Focused on generative modeling and building people-centered AI for communication and expression. |
Work
-
2026.03 - 2026.09 Remote, GB
Computer Vision Researcher Intern
Signapse
Six-month research internship at an AI sign language translation company, working on photorealistic real-time signing avatar rendering and data capture.
- Developed a custom implementation of an open-source 3D Morphable Model with 3D Gaussian Splatting for photorealistic real-time signing rendering, with a focus on human reconstruction
- Designed and built a 3-camera synchronized global-shutter capture rig, shipped abroad to record new sign language datasets, capturing over 30,000 glosses to date
-
2023.09 - Present Guildford, GB
Graduate Teaching Assistant
University of Surrey
Part of the team delivering labs for CS and EEE courses, providing lab instruction and technical mentoring.
- Delivered four lab series to cohorts of 50-200 students, including COM1031 (Computer Logic), EEE1035 (C Programming), and EEE2048 (Algorithms)
- Completed GTA training to enhance teaching, feedback, and teamwork skills
-
2022.06 - 2022.09 Guildford, GB
Summer Research Intern
University of Surrey
Research on multi-agent reinforcement learning strategies for UAV access control, with a focus on reproducibility and model efficiency.
- Reimplemented baselines from Matlab to PyTorch for reproducible experimentation and up to 50% faster training
- Designed and trained deep policy networks using DQN
- Evaluated agent coordination strategies in simulated environments
-
2021.06 - 2022.06 London, GB
Associate Development Engineer
Sky Group
Collaborated with an agile engineering team to develop computer vision tools for automated set-top box validation, integrating OCR pipelines into production workflows.
- Built OCR pipelines for detecting on-screen text and anomalies in video streams
- Deployed computer vision tools for regression testing in content validation workflows
Education
-
2023.09 - 2027.09 Guildford, GB
PhD
University of Surrey
Computer Vision
- Thesis: Translation and Production of Non-Manual Features for Sign Language
- Supervised by Prof. Richard Bowden and Dr. Özge Mercanoğlu Sincan
- Reviewer for IEEE FG and BMVC
- ECCV 2026 Doctoral Consortium
-
2019.09 - 2023.06 Guildford, GB
BEng (Hons)
University of Surrey
Electronic Engineering
- Graduated with First Class Honours
- Computer Vision
- Computer Graphics
- C++ and Object-Oriented Design
- Computer Algorithms and Architecture
Publications
-
2026.08.07 M3T: Discrete Multi-Modal Motion Tokens for Sign Language Production
[Accepted BMVC] arXiv preprint
State-of-the-art 3D sign language production with explicit non-manual features: an autoregressive transformer over discrete FSQ-VAE motion tokens on an SMPL-X + FLAME morphable body model, evaluated on How2Sign, CSL-Daily, Phoenix14T, and NMFs-CSL.
-
2026.06.18 SignSparK: Efficient Multilingual Sign Language Production via Sparse Keyframe Learning
[Accepted ECCV] arXiv preprint
Efficient multilingual sign language production from sparse keyframes: conditional flow matching with keyframe-to-pose editing, automatic sign segmentation, and 3D Gaussian Splatting rendering, across British, American, Chinese, and German Sign Language.
-
2025.07.10 VisualSpeaker: Visually-Guided 3D Avatar Lip Synthesis
ICCV Workshops
High-fidelity 3D avatar lip synchronization via a perceptual lip-reading loss: visual speech recognition supervising differentiable 3D Gaussian Splatting renders, supporting accurate mouthings for sign language avatars.
Skills
| Programming Languages & Frameworks | |
| Python | |
| C++ | |
| PyTorch | |
| OpenCV |
| Tools | |
| Docker | |
| Git | |
| LaTeX | |
| Weights & Biases | |
| Linux | |
| SLURM | |
| Unreal Engine |
| Specialized | |
| Computer Vision | |
| Sign Language Production | |
| Generative Modeling | |
| Flow Matching | |
| 3D Gaussian Splatting | |
| Neural Rendering | |
| 3D Morphable Models | |
| Reinforcement Learning |
Languages
| English | |
| Fluent |
| Greek | |
| Proficient |
| German | |
| Proficient |
| Dutch | |
| Proficient |
Projects
- 2025.05 - 2025.08
PokerGym: Multi-Agent Poker Environment
- Custom Texas Hold'em multi-agent reinforcement learning environment (PettingZoo, Gymnasium) with a full PPO training pipeline, reward tracking, and legal action masking
- 2023.09 - 2024.09
Volumetric Sign Capture System
- Low-cost, scalable 7-camera volumetric capture rig, combining Raspberry Pi cameras with a broadcast-grade Blackmagic studio camera and custom synchronization and control software, for multi-view sign language datasets
- 2022.09 - 2023.06
Marker-less Motion Capture for Signing Avatar
- Undergraduate thesis: real-time photorealistic signing avatar driven from monocular video, combining MediaPipe, a custom uplift network, and Unreal Engine Metahumans
Volunteer
-
2023.09 - 2024.06 Guildford, GB
Societies Chair
University of Surrey Students' Union
Oversaw 150+ student-led societies for the Students' Union.
-
2019.09 - 2025.06 Committee Member / Founder
University of Surrey Societies
Founded GameDev Society; committee roles in GameSoc and EARS.
Interests
| Research | |
| Computer Vision | |
| 3D Graphics | |
| Human Motion Synthesis | |
| Sign Language Translation |
| Hobbies | |
| Game Development | |
| Board Games | |
| Climbing | |
| Weightlifting |
References
| Upon Request | |
| Available upon request from supervisors and colleagues. |