James Thewlis

Co-Founder of Unitary

James Thewlis is the Co-Founder and Chief Scientist of Unitary, a startup working to prevent harmful content online with computer vision. Prior to founding Unitary, James completed a PhD in computer vision from the University of Oxford’s Visual Geometry Group supervised by Andrea Vedaldi, during which he collaborated with Facebook AI Research and published at international AI conferences such as NeurIPS.

news

Oct 3, 2023 Unitary raises $15M Series A round led by Creandum
Sep 19, 2023 Language as the Medium: Multimodal Video Classification through text only accepted to the What is Next in Multimodal Foundation Models? ICCV 2023 workshop!
Mar 7, 2023 Unitary Fundraising Announcement - Unitary gets £8m boost for AI social media moderators
Oct 23, 2022 Our paper VTC: Improving Video-Text Retrieval with User Comments published at ECCV!
May 11, 2021 I have been doing some work adding JPEG decoding on the GPU to the Pytorch torchvision library

selected publications

  1. ECCV
    VTC: Improving Video-Text Retrieval with User Comments
    Hanu, Laura, Thewlis, James, Asano, Yuki M, and Rupprecht, Christian
    In ECCV 2022
    arXiv HTML PDF Code
  2. ICCV
    Unsupervised learning of landmarks by descriptor vector exchange
    Thewlis, James, Albanie, Samuel, Bilen, Hakan, and Vedaldi, Andrea
    In ICCV 2019
    arXiv HTML PDF Code
  3. NeurIPS
    Modelling and unsupervised learning of symmetric deformable object categories
    Thewlis, James, Bilen, Hakan, and Vedaldi, Andrea
    In Advances in Neural Information Processing Systems 2018
    HTML PDF
  4. NeurIPS
    Unsupervised learning of object frames by dense equivariant image labelling
    Thewlis, James, Bilen, Hakan, and Vedaldi, Andrea
    In Advances in Neural Information Processing Systems 2017
    arXiv PDF
  5. ICCV
    Unsupervised learning of object landmarks by factorized spatial embeddings
    Thewlis, James, Bilen, Hakan, and Vedaldi, Andrea
    In ICCV 2017
    arXiv HTML PDF