Brain-Inspired Neural Networks for Computer Vision AI
Reduced Training. Uncompromised Performance. Explainability built in.
The Current Problem, Our Solution
AI models require extensive training to achieve high performance, hindering AI from reaching its best potential responsibly and sustainably.
We aim to be part of the solution by supporting developers and organisations building computer vision (CV) applications in overcoming this hurdle through our AI models.
Our CV models are designed with neural network architectures that reduce the amount of training required – by up to 50% – without compromising model performance.
This enables:
● 20-30% lower deployment costs
● 7-10% improved performance
● 1.5-2x faster training and retraining cycles
And because explainability is built in, our models offer greater trust, reliability, and smoother integration – especially in high-stakes applications where understanding AI decisions is critical.
About the Project
Our work focuses on designing novel neural network architectures inspired by functional principles observed in the human brain, such as spiking behaviour, hierarchy, parallelism, and functional localisation.
As part of the ICURe Discover programme (Innovate UK), we are exploring real-world applications and gathering feedback from those building and deploying AI systems.
Prof. Martin Trefzer
Principle Scientific Advisor
Dr. Jessica Dobson
Amrutha R K
Entrepreneurial Lead
Is This Useful for You?
As part of our market discovery programme, we’re exploring whether these models could be delivered through a Model-as-a-Service (MaaS) offering – giving teams access to plug-and-play or adaptable models without the burden of high training costs.
We’d love to hear how you currently develop or access models – and whether our CV models could support or improve your workflow.