Latest News!

August 8 2017:

Fabio was awarded the Horizon 2020 project "SARAS - Smart Autonomous Robotic Assistant Surgeon", on the development of robotic assistant surgeons for laparoscopy.

The team will be in charge of the vision and cognitive modules of the system. The project has a total budget of €4,315,640: Oxford Brookes' share is €596,073. The project's duration is of 3 years. The agreed start date is Mar 1st 2018. The Coordinator is Dr Riccardo Muradore from University of Verona, Italy. Fabio's role will be Scientific Officer (SO) for the whole project, as well as WP Leader.

List of Horizon 2020 projects funded in 2017

In surgical operations many people crowd the area around the operating table. The introduction of robotics in surgery has not decreased this number. During a laparoscopic intervention with the da Vinci robot, for example, the presence of an assistant surgeon, two nurses and an anaesthetist, is required, together with that of the main surgeon teleoperating the robot. The assistant surgeon needs always be present to take care of simple surgical tasks the main surgeon cannot perform with the robotic tools s/he is teleoperating (e.g. suction and aspiration during dissection, moving or holding organs in place to make room for cutting or suturing, using the standard laparoscopic tools). Another expert surgeon is thus required to play the role of the assistant, to properly support the main surgeon using traditional laparoscopic tools as shown in Figure 1.

The goal of SARAS is to develop a next-generation surgical robotic platform that allows a single surgeon (i.e., without the need for an expert assistant surgeon) to execute robotic minimally invasive surgery (R-MIS), thereby increasing the social and economic efficiency of a hospital while guaranteeing the same level of safety for patients. This platform is called solo-surgeon system.

July 24 2017:

The Artificial Intelligence and Vision team, led by PhD student Gurkirt Singh, in partnership with Andreas Lehrmann and Leonid Sigal of Disney Research, has won second place in the latest CVPR2017 Charades Activity Challenge for action recognition, behind DeepMind's TeamKinetics led by Andrew Zisserman, third place for temporal detection. Leaderboard

The Charades Activity Challenge aims towards automatic understanding of daily activities, by providing realistic videos of people doing everyday activities. The Charades dataset is collected for an unique insight into daily tasks such as drinking coffee, putting on shoes while sitting in a chair, or snuggling with a blanket on the couch while watching something on a laptop. This enables computer vision algorithms to learn from real and diverse examples of our daily dynamic scenarios. The challenge consists of two separate tracks: classification and localization track. The classification track is to recognize all activity categories for given videos ('Activity Classification'), where multiple overlapping activities can occur in each video. The localization track is to find the temporal locations of all activities in a video ('Activity Localization').

Method's description

At a high level, our approach consists of two parallel convolutional neural networks (CNNs) extracting static (i.e., independent) appearance and optical flow features for each frame, plus, there is another parallel audio feature extraction stream using Soundnet CNN and scoring done using an SVM. We fuse information from three streams using a convex combination of their respective classification scores to obtain a final result.
We train the overall network using a multi-task loss: (1) Classification: Both streams produce a C-dimensional softmax score vector that is trained using back-propagation with a cross-entropy loss; (2) Regression: In addition to the classification scores, the appearance stream also produces 3-dim. coefficients for each class describing the offset from the boundaries of the current action as well as its overall duration. This network path is trained using a smooth L1 loss.
The audio stream consists of feature extraction using pretrained soundet CNN and SVM classifier to produce classification in sliding window fashion. Audio scores are interpolated to the same frame as other two stream outputs.
We generate frame-level scores at 12 fps. For temporal action segmentation, we fuse the scores of three streams at the frame-level using a convex combination. The weights to each stream can be found by cross-validation on the validation set. Finally, we produce a score vector for 25 regularly sampled frames using top-k mean-pooling in a temporal window around those frames. Frame-level score for each class is the mean of the top-20 frame-level scores of class c in a temporal window of size 40. Similarly, we apply top-k mean pooling on the scores for class c for the entire duration of video to obtain video classification scores. We found that top-k value of 40 works well via cross-validation.

July 16 2017:

The papers

G. Singh, S. Saha, M. Sapienza, P. Torr and F. Cuzzolin, Online Real-time Multiple Spatiotemporal Action Localisation and Prediction

Link to arXiv version

S. Saha, G. Singh and F. Cuzzolin, AMTnet: Action-Micro-Tube regression by end-to-end trainable deep architecture

Link to arXiv version

were accepted for publication at the International Conference on Computer Vision (ICCV 2017), Venice, Italy, October 2017 - the premiere venue for Computer Vision, as part of the ongoing world-leading action detection project at the Artificial Intelligence and Vision group.

July 6 2017:

Fabio was invited to speak at the Fourth Summer School on Belief Functions and Their Applications (BELIEF 2017)

Title of the talk: The statistics of belief functions

Although born within the remit of mathematical statistics, the theory of belief functions has later evolved towards subjective interpretations which have distanced it from its mother field, and have drawn it nearer to artificial intelligence. The purpose of this talk, in its first part, is to understanding belief theory in the context of mathematical probability and its main interpretations, Bayesian and frequentist statistics, contrasting these three methodologies according to their treatment of uncertain data.
In the second part we recall the existing statistical views of belief function theory, due to the work by Dempster, Almond, Hummel and Landy, Zhang and Liu, Walley and Fine, among others.
Finally, we outline a research programme for the development of a fully-fledged theory of statistical inference with random sets. In particular, we discuss the notion of generalised lower and upper likelihoods, the formulation of a framework for logistic regression with belief functions, the generalisation of the classical total probability theorem to belief functions, the formulation of parametric models based of random sets, and the development of a theory of random variables and processes in which the underlying probability space is replaced by a random set space.

June 2017:

Fabio is elected Executive Editor of the Society for Imprecise Probability - Theory and Applications (SIPTA),

The Society for Imprecise Probability: Theories and Applications (SIPTA) was created in February 2002, with the aim of promoting the research on imprecise probability. This is done through a series of activities for bringing together researchers from different groups, creating resources for information, dissemination and documentation, and making other people aware of the potential of imprecise probability models.
The Society has its roots in the Imprecise Probabilities Project conceived in 1996 by Peter Walley and Gert de Cooman and its creation has been encouraged by the success of the ISIPTA conferences.
Imprecise probability is understood in a very wide sense. It is used as a generic term to cover all mathematical models which measure chance or uncertainty without sharp numerical probabilities. It includes both qualitative (comparative probability, partial preference orderings, …) and quantitative modes (interval probabilities, belief functions, upper and lower previsions, …). Imprecise probability models are needed in inference problems where the relevant information is scarce, vague or conflicting, and in decision problems where preferences may also be incomplete.

June 13 2017:

The paper The Total Belief Theorem, authored by Dr Chunlai Zhou and Professor Fabio Cuzzolin, is accepted for publication at Uncertainty in Artificial Intelligence (UAI) 2017

In this paper, motivated by the treatment of conditional constraints in the data association problem, we state and prove the generalisation of the law of total probability to belief functions, as finite random sets.
Our results apply to the case in which Dempster's conditioning is employed. We show that the solution to the resulting total belief problem is in general not unique, whereas it is unique when the a-priori belief function is Bayesian. Examples and case studies underpin the theoretical contributions.
Finally, our results are compared to previous related work on the generalisation of Jeffrey’s rule by Spies and Smets.

Paper submission PDF

October 2016:

Podcast with Risk Roundup: Advances in AI: Human/Non-Human Action and Gesture Recognition Prof. Fabio Cuzzolin, Head of Artificial Intelligence and Vision at Oxford Brookes University, Oxford, United Kingdom participates in Risk Roundup to discuss ''Advances in Artificial Intelligence: Human and Non-Human Gesture and Action Recognition''.

How would we define and describe man-machine or a machine-machine interface and why is it relevant to understanding Artificial Intelligence? Mediator between human (and non-human users) and machines, a man-machine or machine-machine interface, is basically a system that takes care of the entire human-non-human communication process. It is responsible for the delivery of the machine or computer knowledge, functionality and available information, in a way that is compatible with the end-user’s communication channels, be it human or non-human. It then translates the user’s (human or non-human) actions (user input) into a form (instructions/commands) that is understandable by a machine.

When increasingly complex Artificial Intelligence based systems, products and services are rapidly emerging across nations, the necessity for more user friendly man-machine or machine-machine interface is becoming increasingly necessary for their effective utilization, and consequently for the success that they were designed for.

Published on Risk Group:

October 2016:

Fabio has been invited to be a keynote speaker at CSA 2016, the The 2nd Conference on Computing Systems and Applications. The second edition of the Computing Systems and Applications (CSA) conference will take place from December 13 through December 14, 2016. The conference is open for researchers, academics and industry practitioners interested in the latest scientific and technological advances occurring in different fields of computer science. It constitutes a leading venue for students, researchers, academics and industrials to share their new ideas, original research findings and practical experiences across all computer science disciplines.

CSA 2016 will be held in the Ecole Militaire Polytechnique (EMP) located in Algiers; the capital and the largest city of Algeria. This pioneering engineering college is situated in Bordj El Bahri, a lively city lapped by the Mediterranean Sea and facing the well-known Algiers bay. EMP is one of the oldest technical schools for the training of highly-qualified academics in Algeria. Its know-how covers teaching and research activities in the fields of computer science, electrical and mechanical engineering, and chemistry.

Download the Call for Papers at

July 2016:

Fabio is promoted to Professor

July 14 2016:

invited seminar "Belief functions: past, present and future", part of the statistics colloquia at Harvard University, Department of Statistics.

The theory of belief functions, sometimes referred to as evidence theory or Dempster-Shafer theory, was first introduced by Arthur P. Dempster in the context of statistical inference, to be later developed by Glenn Shafer as a general framework for modelling epistemic uncertainty. Belief theory and the closely related random set theory form a natural framework for modelling situations in which data are missing or scarce: think of extremely rare events such as volcanic eruptions or power plant meltdowns, problems subject to huge uncertainties due to the number and complexity of the factors involved (e.g. climate change), but also the all-important issue with generalisation from small training sets in machine learning.

This short talk abstracted from an upcoming half-day tutorial at IJCAI 2016 is designed to introduce to non-experts the principles and rationale of random sets and belief function theory, review its rationale in the context of frequentist and Bayesian interpretations of probability but also in relationship with the other main approaches to non-additive probability, survey the key elements of the methodology and the most recent developments, discuss current trends in both its theory and applications. Finally, a research program for the future is outlined, which include a robustification of Vapnik' statistical learning theory for an Artificial Intelligence 'in the wild'.

Slides in PDF format

July 13 2016:

The paper Deep Learning for Detecting Multiple Space-Time Action Tubes in Videos, led by first author Suman Saha, was accepted for publication at BMVC 2016 Project web site

In this work we propose a new approach to the spatiotemporal localisation (detection) and classification of multiple concurrent actions within temporally untrimmed videos. Our framework is composed of three stages.
In stage 1, a cascade of deep region proposal and detection networks are employed to classify regions of each video frame potentially containing an action of interest. In stage 2, appearance and motion cues are combined by merging the detection boxes and softmax classification scores generated by the two cascades. In stage 3, sequences of detection boxes most likely to be associated with a single action instance, called {action tubes}, are constructed by solving two optimisation problems via dynamic programming.
While in the first pass action paths spanning the whole video are built by linking detection boxes over time using their class-specific scores and their spatial overlap, in the second pass temporal trimming is performed by ensuring label consistency for all constituting detection boxes.
We demonstrate the performance of our algorithm on the challenging UCF101, J-HMDB-21 and LIRIS-HARL datasets, achieving new state-of-the-art results across the board and significantly lower detection latency at test time.

Arxiv paper coming soon

July 1 2016:

The Artificial Intelligence and Vision research group, led by PhD student Gurkirt Singh, has won second place in the latest CVPR ActivityNet Large Scale Activity Detection Challenge. Leaderboard

The ActivityNet Large Scale Activity Recognition Challenge is a half-day workshop to be held on July 1 in conjuction with CVPR 2016, in Las Vegas, Nevada. In this workshop, we establish a new challenge to stimulate the computer vision community to develop new algorithms and techinques that improve the state-of-the-art in human activity understanding. The data of this challenge is based on the newly published ActivityNet benchmark.

The challenge focuses on recognizing high-level and goal oriented activities from user generated videos, similar to those found in internet portals. This challenge is tailored to 200 activity categories in two different tasks. (a) Untrimmed Classification Challenge: Given a long video, predict the labels of the activities present in the video; (b) Detection Challenge: Given a long video, predict the labels and temporal extents of the activities present in the video.

Report in PDF format

January 2016:

Fabio's tutorial "Belief functions for the working scientist" has been accepted for a half-day presentation at IJCAI 2016, the premiere international conference on Artificial Intelligence, which will take place at the Hilton Midtown Hotel, New York City, on July 9-15 2016.

A dedicated web site can be found HERE.

The theory of belief functions, sometimes referred to as evidence theory or Dempster-Shafer theory, was first introduced by Arthur P. Dempster in the context of statistical inference, and was later developed by Glenn Shafer as a general framework for modelling epistemic uncertainty. Belief theory and the closely related random set theory form a natural framework for modelling situations in which data are missing or scarce: think of extremely rare events such as volcanic eruptions or power plant meltdowns, problems subject to huge uncertainties due to the number and complexity of the factors involved (e.g. climate change), but also the all-important issue with generalisation from small training sets in machine learning.

This tutorial is designed to introduce the principles and rationale of random sets and belief function theory to the wider AI audience, survey the key elements of the methodology and the most recent developments, make AI practitioners aware of the set of tools that have been developed for reasoning in the belief function framework on real-world problems. Attendees will acquire first-hand knowledge of how to apply these tools to significant problems in major application fields such as computer vision, climate change, and others. The performance of these approaches will be critically compared with those of more classical regression, classification or estimation methods to highlight the advantage of modelling lack of data explicitly.

Februry 2015:

Fabio was invited at the Oxford Martin School workshop on "Artificial Intelligence and Predictive Modelling" with Garry Kasparov

Fabio was also invited to a private dinner with Garry and other distinguished guests at Balliol College.

When Garry Kasparov visited the Oxford Martin School this week, he came with a strong message about innovation: society has become too risk averse and we are at risk of failing to innovate if investor mindsets don’t change soon. During two lively workshops, the former World Chess Champion debated the future of innovation with 20 researchers from the University of Oxford, Oxford Brookes and industry. He also delivered a lecture to an audience of 440 at the University of Oxford’s Examination Schools. Top of Kasparov’s agenda was the issue of risk aversion and its impact on societal progress. “A fear of uncertainty holds us back from doing things quickly and productively,” he argued in his second workshop. “Just look the airline industry. Planes are getting better in terms of comfort and fuel efficiency but not going faster. Our preference is for comfort over speed. This mentality is reflected in many different areas; we have become a risk averse society.”

September 2014:

Fabio's monograph entitled "Visions of a Generalized Probability Theory" has been published by Lambert Academic Publishing

The theory of evidence (also known as ‘evidential reasoning’, ‘belief theory’ or ‘Dempster-Shafer theory’) is, perhaps, one of the most successful frameworks for uncertainty modelling, and arguably the most straightforward and intuitive approach to a generalized probability theory. Emerging in the late Sixties from a profound criticism of the more classical Bayesian theory of inference and modelling of uncertainty, evidential reasoning has stimulated in the last four decades an extensive discussion on the epistemic nature of both subjective ‘degrees of beliefs’ and frequentist ‘chances’.

Computer vision is a fast growing discipline whose ambitious goal is to equip machines with the intelligent visual skills humans and animals are provided by Nature, allowing them to interact effortlessly with complex and inherently uncertain environments. This Book shows how the fruitful interaction of computer vision and belief calculus is capable of stimulating significant advances in both fields. Novel results on the mathematics of belief functions are developed in response to the issues posed by fundamental vision problems to which, in turn, novel evidential solutions are proposed.

September 2014:

Springer's Lecture Notes in Artificial Intelligence Volume 8764 entitled Belief Functions: Theory and Applications, edited by Fabio, is available online.

Belief Functions: Theory and Applications
Third International Conference, BELIEF 2014, Oxford, UK, September 26-28, 2014. Proceedings
Series: Lecture Notes in Computer Science, Vol. 8764
Subseries: Lecture Notes in Artificial Intelligence
Cuzzolin, Fabio (Ed.)
2014, XVIII, 444 p. 92 illus.

This book constitutes the thoroughly refereed proceedings of the Third International Conference on Belief Functions, BELIEF 2014, held in Oxford, UK, in September 2014. The 47 revised full papers presented in this book were carefully selected and reviewed from 56 submissions. The papers are organized in topical sections on belief combination; machine learning; applications; theory; networks; information fusion; data association; and geometry.

September 26-28 2014:

The Third Edition of the International Conference on Belief Functions was successfully held in St. Hugh's college, Oxford.

BELIEF 2014, the third edition of the series of conferences on the theory and application of belief functions is already over, and it is time to sum up the outcomes of this exciting experience and draw some lessons for the future of the conference and the community at large.

November 2012:

Fabio's monograph on "The geometry of uncertainty" has been conditionally approved by Springer-Verlag's "Information Science and Statistics" series

The book is about the geometry of various mathematical descriptions of uncertainty, known as "imprecise probabilities", proposed in the last forty years as alternatives or competitors to classical probability theory. These objects can be seen as points living in a certain geometrical space: they can therefore be handled by geometric means. The book provides indeed a geometrical language for working with imprecise probabilities.

The reviewers commented that "there is no other book addressing the Dempster-Shafer theory of evidence in such exhaustive detail", "there has not been a detailed study of the geometry of belief functions and as such I believe this book would be a very welcome addition to the literature."

October 12 2012:

Fabio has been awarded one of the Next 10 Awards by the Faculty of Technology, Design and Environment (TDE).

The committee overseeing the 'Next 10 Programme' met recently and supported Fabio’s application. Activities should begin this academic year at a point to be agreed with the HoD. Rachel Harrison has been assigned as mentor for the programme and Fabio will also liaise closely with Nigel Crook.
A PhD student will be engaged as soon as possible in order to provide maximum strategic benefit to the development of the planned research and growth of the area. A key objective will be the future development of a successful and focused team. The student will be expected to contribute to such things as the development of major funding proposals in addition to carrying out a formal programme of related PhD study.

Next 10 is a research accelerator programme, designed to help the top emerging researchers in the Faculty to progress towards professorial status and a leadership position within their discipline. Involves a Ph.D. Studentship. Start date: October 2012.

September 2012:

Fabio has taken on the role of Head of the Artificial Intelligence (formerly Machine Learning) research group.

September 5 2012:

Fabio has been awarded the Outstanding Reviewer Award at the latest British Machine Vision Conference (BMVC2012) in Surrey.

July 2012:

Fabio's student Michael Sapienza has been awarded the Best Poster Prize at the latest 2012 INRIA Summer School on Machine Learning and Visual Recognition, for his poster "Learning discriminative space-time actions from weakly labelled videos".

Current state-of-the-art action classification methods derive action representations from the entire video clip in which the action unfolds, even though this representation may include parts of actions and scene context which are shared amongst multiple classes. For example, different actions involving the movement of the hands may be performed whilst walking, against a common background. In this work, we propose an action classification framework in which discriminative action subvolumes are learned in a weakly supervised setting, owing to the difficulty of manually labelling massive video datasets. The learned sub-action models are used to simultaneously classify video clips and to localise actions in space-time. Each subvolume is cast as a BoF instance in an MIL framework, which in turn is used to learn its class membership. We demonstrate quantitatively that the classification performance of our proposed algorithm is comparable and in some cases superior to the current state-of-the-art on the most challenging video datasets, whilst additionally estimating space-time localisation information.

July 19 2011:

Fabio has been promoted to Reader, effective September 1st 2011.

July 25 2011:

Fabio has been awarded a best poster award for a his poster entitled "Geometric conditional belief functions in the belief space" at the latest ISIPTA'11 Symposium on Imprecise Probabilities.

In this poster we explore geometric conditioning in the belief space B, in which belief functions are represented by the vectors of their belief values b(A). We adopt once again distance measures d of the classical Lp family, as a further step towards a complete analysis of the geometric approach to conditioning. We show that geometric conditional b.f.s in B are more complex than in the mass space, less naive objects whose interpretation in terms of degrees of belief is however less natural.

July 19 2011:

Fabio has received his tenure and his now a Senior Lecturer with the Department of Computing and Communication Technologies, Oxford Brookes University.

February 23 2011:

Fabio has been awarded support for his EPSRC First Grant! This is a two-year, 122 K pound grant which will involve hiring a postdoctoral researcher in year 2.

November 12 2010:

Fabio has been nominated Associate Editor of the IEEE Transaction on Systems, Man, and Cybernetics - Part C!

June 15 2010:

Following the latest Workshop on the Theory of Belief Functions, Fabio has been elected in the Board of Directors of the Belief Functions and Applications Society with 27 votes

Fabio Cuzzolin received the best paper award for the outstanding technical contribution assigned to the paper:

Alternative formulations of the theory of evidence based on basic plausibility and commonality assignments

at the Tenth Pacific Rim International Conference on Artificial Intelligence (PRICAI-08), Hanoi, Vietnam, 15-19 December 2008: URL:

The Pacific Rim International Conference on Artificial Intelligence (PRICAI) is a biennial international event which concentrates on AI theories, technologies and their applications in the areas of social and economic importance for countries in the Pacific Rim. In the past conferences have been held in Nagoya (1990), Seoul (1992), Beijing (1994), Cairns (1996), Singapore (1998), Melbourne (2000), Tokyo (2002), Auckland (2004) and Quilin (2006).

The paper introduces two novel alternative mathematical formulations of the theory of belief functions or "theory of evidence". We prove that the equivalent representations of evidence given by plausibility and commonality functions have the combinatorial structure of sum functions, just like belief functions do, and we compute their Moebius inverses.