Skip to Content

Automated early detection of health and welfare compromises in pigs using video-based automatic behavioural analysis.

PhD project title

Automated early detection of health and welfare compromises in pigs using video-based automatic behavioural analysis.

Outline description, including interdisciplinary, intersectoral and international dimensions 

Changes in pig behaviours are a useful aid in detecting early signs of compromised health or poor welfare. Early detection of problems allows timely intervention, mitigates losses and improves well-being. On a commercial scale, human observation of the subtle changes in behaviour that may indicate early-stage disease or vice is impractical. The aim of this studentship is to develop methods that automatically detect these behavioural changes using inexpensive, non-invasive and scalable technologies such as video analytics.

The starting point will be deep learning-based methods previously developed by the team to capture a number of relevant factors, such as animal posture and the feeding and drinking behaviours, and social interactions in groups of pigs (as opposed to individual pigs).

The aims of the studentship are:

  1. Develop methods for tracking the behaviour of individuals within a group, with or without additional sensors or explicit identification of individual pigs. Video detection methods will be combined with evidential reasoning networks to deduce behaviour patterns between pigs. Generative adversarial networks will be used to assess the robustness of the prediction methods.
  2. Apply these novel methods to the early detection of two significance issues found in pig systems:
  • the digestive disorders associated with the transition from milk to solids at weaning
  • the occurrence of vice, such as tail-biting or flank chewing.

These compromises are widespread within the industry and we anticipate no difficulty in finding commercial pig production units to collaborate with us (this will be managed by the Industry partner).

  1. Establish unsupervised learning models that baseline normal behaviours that become further enriched over time. Use advanced anomaly detection methods to identify behavioural changes occurring in each pen/group of pigs.
  2. Develop a pipeline to give early warning alerts to production supervisors that identify anomalous individuals or groups of pigs.

The challenge of early detection of health and welfare compromises is of global interest, mainly because >70% of pigs across the world are kept in similar management systems. The project thus has the potential to offer solutions applicable globally.

The studentship will be closely aligned with the EU funded project Healthy Livestock, which involves academic and industry collaborators from several EU countries and China, including Zoetis.  In particular, this will involve close collaboration with Dr Kanellos the Director of the international arm of Zoetis, Business Development and Alliances, a US based company which is the largest in global animal health.  Association with the EU project and Zoetis will offer several advantages to the student, including exposure to international Stakeholders.

The student will be expected to present at international conferences and publish in international journals. The project is aiming to organise a Session at the World Conference on Farm Animal Welfare hosted by the EU project partner International Cooperation Committee of Animal/ Welfare. Other international conferences for dissemination include international Computer Vision Conferences such as CVPR, ECCV and BMVC, where presentations on applications of computer vision are encouraged.  The student will be expected and encouraged to present his/her work at these events. Professor Kyriazakis has an exemplary record of publications in International journals that arise from PhD projects (examples include IEEE journals (Transactions on Automation Science and Engineering), Scientific Reports and Biosytems Engineering). The outcomes of this studentship will be suitable for publication in international multidisciplinary journals.


Key words/descriptors



Animal Health & Welfare

Behavioural Analysis

Computer vision

Video analytics

AI reasoning

Fit to CITI-GENS theme(s)

  • Information Technology,
  • Life Sciences

This is an interdisciplinary project that requires collaboration between computer scientists and engineers on the one hand and animal and veterinary scientists on the other. As such it addresses the above objectives of two of the CITI-GENS themes. This builds upon an existing collaboration between ECIT and IGFS in the area of animal welfare, involving international collaboration as well.

Supervisor Information



First Supervisor:   Professor Ilias Kyriazakis                                                        School: Biological Sciences (IGFS)

Second Supervisor:   Dr Paul Miller                                                                      School: EEECS (ECIT GII)

Third Supervisor:  Dr Theo Kanellos                                                                     Company: Zoetis

Fourth Supervisor:  Dr Ali Alameer                                                                       School: Biological Sciences (IGFS)

What costs are associated with the project and how will they be funded?


NB: The COFUND research grant supports the financing of student fees and the salary of the ‘Fellows.’ Additional overheads (e.g. specialist training, equipment) are not provided for

The main costs associated with the project are:

1) Specialist equipment to capture pig behaviours (cameras of varying quality depending on their resolution),

2) Specialist servers for the data management and modification,

3) Any farm compensation for the in-farm pen modifications and disturbance,

4) Outsourcing of behavioural data annotation.


Costs for 1) and 2) will be covered in part by the Industry partner (Zoetis); different cameras and a server are already owned by the collaborative team, as they have been bought for the purposes of a current EU funded project. ECIT also owns a variety of different camera types.

Costs for 3) will be borne by Zoetis.

Costs for 4) will be covered by the current EU funded project. 


Name of non-HEI partner(s)

 Zoetis Ltd and Zoetis UK


Contribution of non-HEI partner(s) to the project:



Access to commercial pig units and provision of specialist equipment for capturing video imaging of groups of pigs

Student training, participation at relevant meetings and placement at Zoetis

In kind contribution of Dr Kanellos’ time for student supervision.

Profile of non-HEI partner


Zoetis Ltd is the world leading pharmaceutical company specialising in animal products and animal health technologies. They will be the non-HEI partner in the project. Their interest lies in the early detection of health and welfare compromises in livestock and the application of early and targeted treatment for individuals and groups of animals. This is expected to reduce input from antimicrobials in livestock systems.

Zoetis Ltd European headquarters have recently moved to Dublin.

The student will spend two periods of 3 months each during the course of their studies. The timing of these placements will be agreed with Zoetis, but they are likely to happen during the first 6 months of the project (Year 1) and during the first 3 months of the final year of the studies. The aim of the first placement will be for the student to become acquainted with the wider context of their studies. The aim of the second placement would be to explore the application of the findings to commercial situation and the potential for exploitation.

Zoetis has acted as the Industrial sponsor to numerous CASE Studentships (BBSRC and EPSRC) and they are experienced in providing in-house a training programme that meets the needs of co-supervised PhD students.

Research centre / School


Faculty of Medicine, Health and Life Sciences – Institute for Global Food Security – School of Biological Sciences

Faculty of Engineering and Physical Sciences - ECIT/CSIT – School of EEECS

Subject area

Animal and Veterinary Sciences

Computer Vision and AI