Improving equine gait events detection with AI
Image produced with ChatGPT-5.
Problem statement
Locomotor injuries are an important problem for equine (horses) health and ultimately welfare. Early detection of locomotion abnormalities through regular monitoring or follow-up during rehabilitation programs at home can be a solution to avoid developing into more serious conditions.
Inertia Technology B.V. has developed an expert equine gait analysis system, Equi-Pro®, which collects and analyses locomotion data in the veterinary context. This system uses inertia measurement units (IMUs) placed on different body parts to precisely quantify motion in horses.
One open challenge is the accurate detection of gait events such as hoof on and hoof off (contacts of the limbs with the ground, see here) during different types of movements, at different speeds and on different surfaces.
Aim
The overall aim of this project is to develop (artificial intelligence) models and algorithms to accurately and precisely detect hoof events from limb- and/or body-mounted IMUs.
Data and baseline algorithms (MATLAB .p files) are provided by the partner company for this assignment.
Tasks
Using provided data and labels, your tasks will be:
- Developing a pipeline to clean the manually defined hoof on/off timestamps (labels) and correct them if needed
- Using the labels as ground truth and the raw or processed IMU signals, develop algorithms/train AI models to automatically detect hoof on/off events
- Developing a benchmark to systematically compare the results of your models to our baseline algorithms
Work
- 20% Theory: Research on equine biomechanics, IMU technology, algorithms and AI models for hoof events detection
- 70% Prototyping and Testing: Develop and test the algorithms/AI models and benchmarking pipeline
- 10% Writing: Summarise the findings, implementation strategies and recommendations for real-world deployments.
While most of the work can be conducted with Python, current implementations are in MATLAB. Therefore, MATLAB is preferred.
Moreover, if the results are encouraging, we could work together towards a scientific publication.
This project will be in collaboration with Inertia Technology B.V.
Contact
FURTHER READING