Discover how AI is transforming image recognition in intellectual property
Download this white paper to discover how AI is transforming intellectual property and the tangible benefits it can bring to your company.
How AI is transforming trademark and logo protection
Since 1973, intellectual property offices around the world have used the Vienna Classification to organise and compare registered logos. For decades, this classification and prior art search work relied primarily on manual coding by examiners — a time-consuming, subjective process prone to error. Coexya has been supporting these offices for over 25 years with specialist software solutions, including Acsepto, a suite dedicated to searching for verbal and figurative similarities in registered trademarks.
Five stages of technological evolution over twenty-five years
This white paper traces the history of Acsepto by Coexya from its first version in 1998 through to current deep learning-based technologies. The first version relied on Vienna code searches using logical operators. In 2009, Acsepto became the first trademark similarity search software to combine the Vienna Classification with an independent automatic image recognition module — an advance immediately welcomed by users. From 2010 onwards, the R&D teams tested and combined six distinct algorithms. In late 2016, the R&D division integrated supervised deep learning via convolutional neural networks, drawing on training datasets built with eight partner intellectual property offices. Since 2018, the solution has offered combined concept, shape, texture and colour recognition — capable of comparing two million images in under two seconds.
An R&D partnership with the Cnam to go further
Since 2017, Coexya has collaborated with researchers at the Cnam (Conservatoire National des Arts et Métiers) to guide its machine learning research applied to intellectual property. This collaboration gave rise to a jointly supervised PhD thesis entitled “Robust image retrieval with deep learning”, whose findings were published at three prestigious scientific conferences.
In production at intellectual property offices worldwide
The Swedish Patent & Registration Office uses Acsepto on a daily basis. Anders Svensson, Product Owner, shares his experience: before Acsepto, the office used two separate systems and had to wait 24 hours for results. Acsepto consolidated all searches into a single tool with instant results. The AI module handling figurative trademark classification has even been given an internal nickname — Acke. Current clients of Coexya’s SIP Business Unit include New Zealand, Norway, Ireland and South Africa.
A publication by the experts of Coexya’s SIP Business Unit:
Sylvain Cauchy, SIP Division Director — Benoit Ehrenmann, Account Manager — Xavier Bitot, Head of Expertise
