close
CVC News

AllRead, a Computer vision Centre spin-off, devises a novel technology for the detection and digitisation of images

WhatsApp Image 2019-12-03 at 10.40.10
The five cofounders of AllRead - Meritxell Bassolas /Computer Vision Centre

Researchers from the Computer Vision Centre (CVC-UAB) have created AllRead Machine Learning Technologies (AllRead MLT) a new spin-off specialized in detecting and digitising text in images. Its aim: to push into market a singular technology based on Computer vision and Deep Learning techniques, which can be applied to the identification and tracking of containers in both ports and airports.  

This disruptive solution offered by AllRead allows the localization of a numerical text or code in any part of a given image without the need of segmenting it (the image). What’s more, as explained by its founders, it is easier to implement and use than other commercial solutions.

This new technology is based in a neural network model which can be trained and adapted to different operative or industrial scenarios. The algorithms designed are able to learn from mistakes and improve, making data extraction highly precise. “We process the images and extract the relevant information, that is, the data which has value for our clients. All this process will gain speed when we implement the 5G network”, as explained by Miguel Silva-Costenla, AllRead cofounder and CEO.

Pilot study with the Port of Barcelona

The spin off has already undertaken a number of pilot tests in collaboration with the Port of Barcelona as well as with companies from different sectors, such as Suez, Comsa or International Airlines Group. In all cases, AllRead technology has allowed the improvement of logistic processes by reducing drastically the time span of the extraction of information process, making the operation quicker and more precise. “With a simple infrastructure, that is, a security camera and our software, we can read and extract the information from port containers or water meters”, states Silva-Constenla. This artificial vision solution can be adapted to any data extraction scenario in which the process of reading is crucial: serial numbers, barcodes, sensor reading, expiry dates, etc.

From the UAB to the Mobile World Capital

The AllRead MLT Project started in 2016, thanks to the Idea Generation Program organised by UAB’s Research Parc. The definitive boost came with the participation in the programme The Collider from the Mobile World Capital Barcelona. This programme is intended to foster collaboration between different agents, and was the contact point for three Computer Vision Centre researchers, doctors Dimosthenis Karatzas, Marçal Rossinyol and Lluís Gómez, who were working on an intelligent reading system with two entrepreneurs, who are now scaling the business, Miguel Silva-Constenla and Adriaan Landman.

On March of this year, they accomplished an investment of 50kc, the main investor being the Mobile World Capital. Their aim now is to close new funding during year 2020.

alexandra canet

The author alexandra canet