772-222-5577 operations@shiptms.com

This article is an extract from the Summer Edition of MER Magazine published in April 2019. You can read the full article as well as other articles from MER for free by visiting digital.mailandexpressreview.com.

Lorenzo D’Arsiè, R&D Engineer at Prime Vision, explains how his team used machine learning to tackle the industry-wide problem of incorrectly formatted address labels.

People have an irrational fear of AI. Whether it’s a terminator-led apocalypse, a dumbing-down of mankind, or merely robotic workforces causing unemployment on a global scale; in general, opinions are negative.

The reality is that for the foreseeable future, we are very much in control. We set the rules and parameters under which these machines operate. They are brilliant but only across a very narrow band. The human mind, however, is unfathomably complex. Processing occurs across an incredibly broad spectrum, allowing people to touch, assess, think, plan, calculate, articulate and co-ordinate movement simultaneously.

To put this in perspective, for engineers and developers working at the cutting-edge of AI, to teach a machine to recognise a passive object, assess it and then grasp is a remarkable thing to do. A baby can do this at four months old.

For me, as an R&D engineer at Prime Vision, AI is an incredibly exciting technology. But, like a baby, AI is just talking its first intrepid steps into a new world. A world
with endless possibilities offering enormous potential and benefits for all.

Sometimes it’s hard, not to view AI, and in particular machine-learning projects in terms of a parent-child relationship. We take great pride in new abilities and skills learned. We watch them mimicking their teachers, picking up their habits and shortcuts. Learning how to best approach a problem, and through trial and error, learn to overcome it. They benefit from mistakes, applying reason and logic to think through a process and achieve a defined solution.

Labels: a very traditional problem

Prime Vision’s Chinese Address Reader, launched in October 2018, is an excellent example of AI driving our latest innovations. With mainstream Optical Character Recognition (OCR) machines continually struggling to read incorrectly formatted labels originating from Asia, our working solution promises to set a new standard for OCR techniques.

In its most basic form, the problem is a reciprocal lack of understanding of respective address formats across the Asian and Western logistics Industries. No-one is to blame. It stems from the significant cultural gap that occurs as these markets swiftly overlap.

The explosive growth of the e-commerce market gave a surprising lifeline to the flagging national postal services. Few predicted the market erupting as it did, and establishing itself as an essential global industry servicing the unquenchable appetite of the internet. Initially, the industry reacted well and newly commissioned systems coped with a steady flow of parcels. But the sudden emergence of the Chinese e-commerce market and in particular Alibaba has been a total game-changer and placed new, immense pressures of already stretched infrastructure.

Before the internet, and during the early stages of e-commerce, products were still shipped to a western supplier via a container…

This article is an extract from the Summer Edition of MER Magazine published in April 2019. You can read the full article as well as other articles from MER for free by visiting digital.mailandexpressreview.com.