The Sullivan Doctoral Thesis Prize

The Sullivan Doctoral Thesis Prize celebrates the best PhD thesis in computer vision completed at a UK university. The British Machine Vision Association (BMVA) funds the prize each year in memory of the late Professor Geoff Sullivan. Awards are presented at the British Machine Vision Conference (BMVC).

Apply through the CMT submission page by 17 September 2027. Full details of what to send are below.

Questions

If you have any questions, or would like to talk through whether your thesis is suitable, we are happy to hear from you. Please email the Sullivan Thesis Prize Chairs at Dr Alex Mackin.

Who can apply

You can apply if you passed your PhD viva at a university in the United Kingdom between 1 September 2026 and 31 August 2027. We use the date of your successful viva, shown on the signed examiners’ reports. Your thesis can be on any topic covered by the current BMVC call for papers.

Prizes

The winner receives £1,000 and free registration for BMVC to attend and collect the award in person. We also give prizes of up to £500 to theses that are strong in a particular area, such as their originality or their real-world impact.

The winner writes a short plain-language summary of their work, about 200 words, which we publish on this page. Commended candidates are welcome to do the same.

Every thesis that is submitted is added to the BMVA’s public thesis archive, a lasting and citable record of your work. We also endeavour to give every entrant a short note of feedback on their thesis.

How we choose

The Sullivan Thesis Prize Chairs consider every thesis and judge its overall quality. The list below shows the qualities we look for. It is a guide, not a scoring formula, and each Chair weighs the qualities in their own way.

We publish the list so you know what to expect. It is not a set of targets to write towards. We judge the research itself.

What we look for What it means
Originality and novelty Genuinely new ideas, methods, or framing, not a derivative extension
Significance and impact How much the problem matters and how far the work influences the field or real-world practice
Rigour and soundness Correct, careful method; fair baselines; honest reporting
Reproducibility and openness Released code, data, and models, and results others can verify
Contribution to the field A lasting advance in the field
Clarity and presentation A clear, well-argued account of the problem and the findings
Breadth and scope Engagement with prior work and an appropriate scope for the thesis

Each Chair reads and judges the theses alone first. We then discuss them together and agree the result. We tell every applicant the outcome before BMVC.

Using AI

You may use AI tools, but the main ideas, analysis, and conclusions must be your own. Tell us how you used AI, and include a signed declaration that the work is your own. What matters is what the AI did, not which tool you used:

We do not use AI detectors. We rely on your declaration, your supervisor’s statement, the examiners’ reports, and whether your results can be reproduced.

What to send

Please submit the following through CMT:

  1. A summary of your thesis, at most two pages. Explain the main ideas, the main technical results, why the work matters, and any real world uses. List your main publications (no more than ten) and, for any with co-authors, say what you did.
  2. Your full PhD thesis as a PDF, or a link if the file is large.
  3. Proof that you passed your viva. Send either a recommendation letter from your external examiner, or both the pre viva and post viva reports.
  4. Your AI statement and declaration (see “Using AI” above).

You may also include a short statement from your supervisor, but this is optional. A report from your internal examiner is also welcome, but not required.

Please send all files as PDFs, or give a link we can open. If any file is confidential, please get permission before you send it. By applying, you agree that the BMVA may host your thesis in its public thesis archive.

Apply here: CMT submission page

Deadline: 17 September 2027

Photos of past winners

Thesis Prive Awarding

Thesis Prive Awarding

Thesis Prive Awarding

Thesis Prive Awarding

Past winners

Year Winner Institution Thesis
2026Subhadeep KoleyUniversity of SurreyDemocratising Sketch-based Vision Applications via Foundation Models
2025Alberto PepeUniversity of CambridgeMachine Learning with Geometric Algebra: Multivectors for Modelling, Understanding and Computing
2024Liliane MomeniUniversity of OxfordSign Language Understanding using Multimodal Learning
2022Shangzhe WuUniversity of OxfordUnsupervised Learning of 3D Objects in the Wild
2021Jaime Spencer MartinUniversity of SurreyLearning Generic Deep Feature Representations
2020Dimitrios KolliasImperial College LondonAffect Recognition & Generation in-the-wild
2019Alex KendallUniversity of CambridgeGeometry and Uncertainty in Deep Learning for Computer Vision
2018Oscar MendezUniversity of SurreyCollaborative Strategies for Autonomous Localisation, 3D Reconstruction and Pathplanning
2017Karel LebedaUniversity of Surrey2D and 3D Tracking and Modelling
2016Xiatian ZhuQueen Mary, University of LondonSemantic Structure Discovery in Surveillance Videos
2015Vibhav VineetOxford Brookes UniversityRecognition, Reorganisation, Reconstruction and Reinteraction for Scene Understanding
2014Mattias HeinrichUniversity of OxfordDeformable lung registration for pulmonary image analysis of MRI and CT scans
2013Patrick Ott
(joint winner)
University of LeedsSegmentation Features, Visibility Modeling and Shared Parts for Object Detection
2013Shoaib Ehsan
(joint winner)
University of EssexImproving the Effectiveness of Local Feature Detection
2012Marco PaladiniQueen Mary, University of LondonDeformable and Articulated 3D Reconstruction from monocular video sequences
2011Charles BibbyUniversity of OxfordProbabilistic Methods for Enhanced Marine Situational Awareness
2010Olly OechsleUniversity of EssexTowards the Automatic Construction of Machine Vision Systems using Genetic Programming
2009Pawan Kumar MudigondaOxford Brookes UniversityCombinatorial and Convex Optimization for Probabilistic Models in Computer Vision
2008Pushmeet KohliOxford Brookes UniversityMinimizing Dynamic and Higher Order Energy Functions using Graph Cuts
2007Josef SivicUniversity of OxfordEfficient visual search of images and videos
2006Rob FergusUniversity of OxfordVisual Object Category Recognition
2005Björn StengerUniversity of CambridgeModel-based hand tracking using a hierarchical Bayesian filter
2004Jonathan StarckUniversity of SurreyHuman Modelling from Multiple Views
2003Rhodri DaviesUniversity of ManchesterLearning Shape: Optimal Models for Analysing Shape Variability
2002Albert ChungUniversity of OxfordVessel and aneurysm reconstruction using speed and flow coherence information in phase contrast magnetic-resonance angiograms.
2001Gareth J EdwardsUniversity of ManchesterLearning to identify faces in images and sequences.
2000Richard BowdenBrunel UniversityLearning non-linear Models of Shape and Motion
1999Neil JohnsonUniversity of LeedsLearning Object Behaviour Models

Note: The 2023 prize was not awarded.