Oxford’s Visual Geometry Group, led by professors Andrew Zisserman and Andrea Vedaldi, will work with Nielsen’s Image Recognition Group to more precisely and quickly identify in-store products based on Nielsen’s product reference data.

The researchers will build on the eCollection tool’s algorithms using advanced deep learning, which Nielsen expects will improve accuracy and reduce the costs of its product identification and classification.

Arun Ramaswamy, CTO, Nielsen Global Connect, said: “Artificial intelligence has been extremely successful in using the scale of big data to drive value — but the next step is activating the very small details — like correctly and automatically recognising specific features on different products, even when those products have similar visual characteristics.” 

Vedaldi, associate professor of engineering science, University of Oxford, added: “Our collaboration with Nielsen provides us with the breadth of data to leverage the power of deep learning and to drive new innovations in the area, as we continue to work toward advancing neural networks and image understanding.”