research

2025

1. 

   Neuralatex: The World’s First Machine Learning Library Written Entirely in LaTeX

   Will Rowan

   arXiv preprint arXiv:2503.24187, 2025

   Abs Bib HTML

   We present Neuralatex, the first machine learning library written entirely in LaTeX. When compiled, it generates training data, trains neural networks, runs experiments, and produces figures, all within the LaTeX compilation process. Includes implementations of MLPs, CNNs, ResNets, Transformers, and MiniGPT.

   @article{rowan2025neuralatex,
     title = {Neuralatex: The World's First Machine Learning Library Written Entirely in LaTeX},
     author = {Rowan, Will},
     journal = {arXiv preprint arXiv:2503.24187},
     year = {2025},
   }

2024

1. 

   N Heads Are Better Than One: Exploring Theoretical Performance Bounds of 3D Face Reconstruction Methods

   Will Rowan, Patrik Huber, Nick Pears, and 1 more author

   In ECCV Workshop on Foundation Models for 3D Humans, 2024

   Abs Bib

   We explore upper bounds on 3D face reconstruction accuracy using multiple views and establish benchmarks for single-view methods.

   @inproceedings{rowan2024nheads,
     title = {N Heads Are Better Than One: Exploring Theoretical Performance Bounds of 3D Face Reconstruction Methods},
     author = {Rowan, Will and Huber, Patrik and Pears, Nick and Keeling, Andrew},
     booktitle = {ECCV Workshop on Foundation Models for 3D Humans},
     year = {2024},
   }

2. 

   How Many OptiFaces? A New Evaluation Metric for 3D Face Reconstruction

   Will Rowan, Patrik Huber, Nick Pears, and 1 more author

   In International Conference on Learning Representations (ICLR), 2024

   Abs Bib

   We propose OptiFaces, a new evaluation metric that addresses limitations of existing 3D face reconstruction benchmarks by measuring how many optimised face shapes are needed to represent a dataset.

   @inproceedings{rowan2024optifaces,
     title = {How Many OptiFaces? A New Evaluation Metric for 3D Face Reconstruction},
     author = {Rowan, Will and Huber, Patrik and Pears, Nick and Keeling, Andrew},
     booktitle = {International Conference on Learning Representations (ICLR)},
     year = {2024},
   }

3. ECCV-W

   San Vitale Challenge: Automatic Reconstruction of Ancient Colored Glass Windows

   Nicolo Di Domenico, Guido Borghi, Annalisa Franco, and 2 more authors

   In ECCV Workshop on Artificial Intelligence for Digital Humanities, 2024

   Bib

   @inproceedings{sanvitale2024,
     title = {San Vitale Challenge: Automatic Reconstruction of Ancient Colored Glass Windows},
     author = {Di Domenico, Nicolo and Borghi, Guido and Franco, Annalisa and Rowan, Will and others},
     booktitle = {ECCV Workshop on Artificial Intelligence for Digital Humanities},
     year = {2024},
   }

2023

1. 

   Text2Face: 3D Morphable Faces from Text

   Will Rowan, Patrik Huber, Nick Pears, and 1 more author

   In International Conference on Learning Representations (ICLR), 2023

   Abs Bib HTML

   We present Text2Face, the first method for generating 3D face shape from text descriptions using a 3D Morphable Model.

   @inproceedings{rowan2023text2face,
     title = {Text2Face: 3D Morphable Faces from Text},
     author = {Rowan, Will and Huber, Patrik and Pears, Nick and Keeling, Andrew},
     booktitle = {International Conference on Learning Representations (ICLR)},
     doi = {10.48550/arXiv.2303.02688},
     year = {2023},
   }

2022

1. 

   The Effectiveness of Temporal Dependency in Deepfake Video Detection

   Will Rowan, and Nick Pears

   arXiv preprint arXiv:2205.06684, 2022

   Bib HTML

   @article{rowan2022deepfake,
     title = {The Effectiveness of Temporal Dependency in Deepfake Video Detection},
     author = {Rowan, Will and Pears, Nick},
     journal = {arXiv preprint arXiv:2205.06684},
     year = {2022}
   }