Toby Perrett

prof_pic.jpg

Hi! I’m a senior postdoctoral researcher at the University of Bristol, working with Professor Dima Damen, and co-advising Chiara Plizzari and Saptarshi Sinha. I’m a founding member of the team collecting the EPIC Kitchens datasets, and am also working on the Visual AI project led by Professor Andrew Zisserman.

My research is on video understanding, particularly from egocentric cameras. I’m interested in improving models and representations when labelled data is scarse or imbalanced, and exploring how to generalise to unseen domains. I’m currently looking at how we can exploit 3D and long-term temporal knowledge to improve reasoning in these situations.

news

Apr 9, 2024 New paper on arXiv: Spatial Cognition from Egocentric Video: Out of Sight, Not Out of Mind. Web. Video. PDF. Code coming soon.
Sep 30, 2023 I was an Outstanding Reviewer at ICCV 2023!
Sep 29, 2023 I gave a talk at the JADE (UK Supercomputing Facility) 2023 event. We discussed why egocentric vision is important, and the computational requirements of egocentric datasets and video understanding models.
Jul 14, 2023 Paper accepted at ICCV 2023! What can a cook in Italy teach a mechanic in India? Action Recognition Generalisation Over Scenarios and Locations. Code, benchmarks and models available.
May 15, 2023 I gave a talk at Samsung AI Centre Cambridge. We looked at the difficulties of scaling few-shot models to handle long-tail tasks.
Feb 10, 2023 I gave a talk at the University of Exeter Computer Science Seminar Series. This included a brief history of image and video datasets, how their properties can cause models to take shortcuts, and some recent solutions.
Feb 3, 2023 Paper accepted at CVPR 2023! Use Your Head: Improving Long-Tail Video Recognition. Code, benchmarks and models available.
Jan 27, 2023 I gave a talk at the Visual AI group at the University of Oxford on our latest long-tail video work.

selected publications

2024

  1. osnom.jpg
    Spatial Cognition from Egocentric Video: Out of Sight, Not Out of Mind
    Chiara Plizzari, Shubham Goel, Toby Perrett, Jacob Chalk, Angjoo Kanazawa, and Dima Damen
    In ArXiv, 2024

2023

  1. lmr.jpg
    Use Your Head: Improving Long-Tail Video Recognition
    Toby Perrett, Saptarshi Sinha, Tilo Burghardt, Majid Mirmehdi, and Dima Damen
    In CVPR, 2023

2022

  1. rescaling.gif
    Rescaling Egocentric Vision: Collection, Pipeline and Challenges for EPIC-KITCHENS-100
    Dima Damen, Hazel Doughty, Giovanni Maria Farinella, Antonino Furnari, Evangelos Kazakos, Jian Ma, Davide Moltisanti, Jonathan Munro, Toby Perrett, Will Price, and Michael Wray
    IJCV, 2022

2021

  1. trx.jpg
    Temporal-Relational CrossTransformers for Few-Shot Action Recognition
    Toby Perrett, Alessandro Masullo, Tilo Burghardt, Majid Mirmehdi, and Dima Damen
    In CVPR, 2021

2019

  1. ddlstm.jpg
    DDLSTM: Dual-Domain LSTM for Cross-Domain Action Recognition
    Toby Perrett, and Dima Damen
    In CVPR, 2019

2018

  1. rescaling.gif
    Scaling Egocentric Vision: The EPIC-KITCHENS Dataset
    Dima Damen, Hazel Doughty, Giovanni Maria Farinella, Sanja Fidler, Antonino Furnari, Evangelos Kazakos, Davide Moltisanti, Jonathan Munro, Toby Perrett, Will Price, and Michael Wray
    In ECCV, 2018