A Spanish Start-up is disrupting existing OCR (Optical Character Recognition) systems and bringing major speed potential to multimodal transport processes with its Deep Neural Network (DNN) reading solution. CargoForwarder Global spoke to Adriaan Landman (AL), COO and Co-Founder of AllRead MLT (Machine Learning Technologies), about the company’s developments and its experiences during the COVID-19 crisis.
Speed is of the essence – especially now, during the crisis – where cargo capacity and resources are scarce, budgets strained, and visibility in a largely fragmented process is crucial. Knowing where a container is, what is inside it, where it needs to go and when, is gold. The faster this information can be accessed, the greater the gains: in through-put time, in resource savings and ultimately, in CO2 output reduction – as is the case when trucks do not have to wait in line for longer than absolutely necessary. The positive environmental impact is clearly one that Adriaan Landman is passionate about. That, and the social aspect of providing sustainable, comparatively low-cost, flexible solutions that interact with existing hardware set-ups, and pack a major punch in efficiency potentials.
So what does AllRead MLT do exactly?
AL: “We offer computer-vision-based technology which can read and interpret images. Three scientists at the Computer Vision Center (CVC) of Catalonia (Autonomous University of Barcelona – UAB) developed a Deep Neural Network model that processes images in real-time, with specific features able to read alphanumeric texts at a level of accuracy that is disrupting existing OCR technologies and benchmarks. It uses Deep Learning and rapidly locates and reads text anywhere in an image, at any angle, regardless of dust, blurring, stains, rotation, and movement. The technology is very robust and therefore reaches levels of accuracy that are very interesting for operational environments such as in the cargo industry. The world is imperfect – text, tags, labels, or printed information on containers, for example, get damaged with time, by weather or movement, and become difficult to read. This is where our technology helps – very simply and using existing infrastructures such as security cameras or smartphones. It is completely adaptable to existing hardware and software”
Colliding brains with business expertise results in a top start-up
In DEC19, Spanish entrepreneur and startup newspaper, El Referente, ranked AllRead MLT as Spain’s 7th most innovative startup in its Top 100 rating. Located in Barcelona, Europe’s 6th largest innovation hub and home to the Mobile World Congress – the most significant event in the mobile technology industry and the first global conference to fall victim to COVID-19 this year – AllRead MLT is a result of The Collider Venture Building program (run by Mobile World Capital Barcelona), which “collides the university academic world with the business and entrepreneurship world” and thus matched the three CVC scientists, Dimosthenis Karatzas, Marçal Rossinyol, and Lluís Gómez, with two experts in innovation projects, Miguel Silva-Constenla and Adriaan Landman. Launched in MAR19, the team of five, which has since expanded to ten and recently, in APR20, raised a half-million-euro seed round, is targeting a 600,000-euro turnover this year, focusing on the utilities, manufacturing, and logistics sectors – in particular those in air cargo logistics and at sea ports.
No need for the gasman to call – all you need is a smartphone
AL: “Our mission is to democratize intelligent reading systems in supply chains around the world, to improve operational efficiency and data management.” Initially developed a few years ago as a response to a specific use case: how to get accurate gas meter readings in a simple and cost-effective way, the DNN technology was fed with millions of images of different gas meters and today anyone with a smartphone can simply take a photo of their meter while connected to their account and be billed accordingly. A solution that removes the need for huge investments in changing a nation’s gas meters to fit a particular homogenous, modern smart-meter standard, or sending out employees to take readings in a time-consuming process.
Successfully navigating IAG’s Hangar51 2019/2020
AL: “While we continue our work with the utility sector and are now also working on water meters, we discovered in our talks with industry representatives, that the message of the market was particularly strong from the logistics sector. Supply chain logistics, despite being in a world of virtual reality, digital twins and big data, is still largely focused on text – tons of different forms, labels, stickers, painted or engraved serial numbers, license plates, etc. – all relevant data that companies spend a lot of money on in logging or extracting to identify where their particular assets are at any one point along the chain, often requiring a great deal of error-prone and time-intensive manual labor, and using a multitude of limited legacy systems.”
On 03SEP19, AllRead MLT was one 13 finalists chosen out of 474 applications from 50 countries, to participate in IAG’s Hangar51 accelerator program which specifically matches start-ups to address problems that have been expressed by the Group’s business units. AllRead MLT was selected for the “future cargo logistics” category and had 10 weeks of intense collaboration with IAG Cargo to work on improving IAG Cargo’s ULD management. The DNN algorithm was trained to correctly recognize ULD containers and read container numbers from thousands of ULD images. At the Demo Day on 23JAN20, AllRead MLT demonstrated an 80% efficiency improvement in air cargo container inventory when compared to the existing, largely manual process in place, which “if you extrapolate that to the number of hours of people involved in the process, then you can really appreciate an improvement in quality, in cost reduction, and better allocation of resources.” Since COVID-19 hit just as plans for implementation were being developed, both parties are currently hoping to resume the project once this extraordinary situation is over.
Optimizing the control and monitoring of cargo traffic at ports
"The Port of Barcelona is interested in having more and better information about the containers coming and going by rail, and their position in the convoy. Our technology, applied to all rail transits in the port, would improve monitoring, accelerate the unloading process, and could foster the transition to rail transport," CEO Silva-Constenla outlines another of AllRead MLT’s initial projects. Adriaan Landman explains that there are more than 100 million sea container transits in EU ports every year (ref. EU data 2018), with the majority going on via road feeder services, so “if road transport accounts for 75% of the total inland freight transport – 75 million truck transits – every minute of the average waiting time represents a lot of fuel burn and impact on the environment. Our solution looks at operational efficiency, reducing costs, and improving customer service because less waiting time means less per hour trucking costs, less gasoline, and a reduction of CO2.” Meanwhile, the company is working with the ports of Barcelona, Bilbao, and Algeciras, with others in the pipeline, increasing efficiencies and transparencies over what is passing through the port and from where. Using the existing video camera set-up, the technology reads containers, truck license plates, trailer license plates, dangerous goods labels and gives information on vehicle types passing through the port. All the information is intelligently combined, helping to track, monitor, and identify container movements, and speed up transits.
The future is limitless – and unstoppable
As Marcal Rossinyol, CTO, Co-Founder, and one of the brains behind the technology, puts it: "Machine learning, as the new industrial revolution, is changing the economy towards one dominated by automated decision processes." AllRead MLT has many plans in line with potentials slowly starting to emerge – for example, using drones in warehouses that, equipped with its DNN, will be able to read even damaged barcodes from any angle – a major improvement on the restricted laser-scanning taking place today, where the barcode needs to be intact and placed just right for the scanner to pick it up. And while AllRead MLT is currently cooperating with airlines and ports, there is huge potential also for rail operators, freight forwarding and ground-handling companies – not to mention all the emerging digital ecosystems.
AL: “Our goal is to streamline logistic processes that require rapid and accurate tracking and identification of assets, using computer vision and machine learning techniques to detect and digitize any alphanumeric text.” Basically any process where data reading or the keying in of data is involved, has DNN potential, and Landman points to quick project implementation times: “We always start with a pilot project of around 1-2 months, where we demonstrate in a controlled environment, that our technology solves the problem. Once validated, we roll out. Depending on the complexity of the project, it then takes 2-4 months of implementation, or even quicker.”
What has been positive about corona?
AL: “While it is true that we are building less collective intelligence because of remote work, and therefore hope to get back to the office soon, at least for a day or two each week, I can say that the collaboration methods in our team have incredibly improved. We are more efficient, focused, and punctual in our online meetings. I have recruited 3 people since we are in confinement. They feel fully involved and fully integrated in the team despite never having seen the other people in real life. The use of project management tools, of digital planning and scrum-based tools has really improved and will have a great impact for us also in future. And regarding the market, everybody is way more sensitive on the topic of digitalization and people are more willing to listen and move forward. In the case of the Port of Algeciras, we were able to successfully finalize the project remotely, in collaboration with the onsite team.” He philosophically points to the corona pandemic also disrupting us personally, forcing us to change our thought processes and adapt to new approaches.
That need for adaptation to new approaches is visible in the many air logistics webinars that CFG has attended over the past three months and, as AllRead MLT’s CEO and Co-Founder, Miguel Silva-Constenla confirms, is crucial to business success: "Covid19 will accelerate the digitalization process worldwide across every single corporation. It's not an option anymore for companies to wait and get this thing done, it has become a survival issue."
We always welcome your comments to our articles. However, we can only publish them when the sender name is authentic.