Helping quick digitalization of industries and supply chains, Allread MLT is disrupting traditional OCR with its computer vision technology and neural networks
Small Wonder, an American comedy sitcom that aired in the 1980s showing a household robot reading a book in a matter of seconds just by looking at the pages, is not a sci-fi invention anymore. Barcelona-based deep tech startup Allread MLT has turned that into reality with a software that can extract information at a glance from images captured with mobile phones or cameras.
Founded in March 2019, Allread MLT's software can read and convert all types of text, symbols or codes that appear in industry and supply chain processes into big data, so its corporate users can identify and monitor a variety of products, vehicles or devices.
Traditional OCR technology detects character by character and so requires text to be perfectly framed and in good condition. However, Allread MLT’s SaaS platform is trained to distinguish and extract text automatically, bypassing irrelevant content with precision, minimizing the risk of error.
The technology, fully patented and integrated into Allread MLT software, is the result of five years of investigation in the field of computer vision. Its AI is developed on convolutional neural network (CNN) architecture, a deep learning neural network mostly applied to visual imagery imitating human brain when processing data and creating decision making patterns.
"We don’t want to represent another technological advance in the sector; we want to disrupt the entire operational model” said CEO Miguel Silva-Constenla. “With infrastructure as simple as a security camera and software, you can read all the information, from containers to water bottles,” he added.
Meeting of minds
A spin-off from the Computer Vision Center (CVC) and the Autonomous University of Barcelona, Allread MLT was founded when a couple of computer scientists got together with a pair of innovation disruptors to apply advances in the field of programming and computer engineering to automate tasks performed by humans.
Dimosthenis Karatzas and Marçal Rossinyol, both CVC researchers with significant experience in computer vision technology and document image analysis, teamed up with serial entrepreneur Silva-Constenla and operations expert Adriaan Landman.
Rossinyol had participated in over 20 research projects for the European Commission while Karatzas won an award in 2013 at the International Conference on Document Analysis and Recognition for his outstanding service and innovative research in human perception-based document analysis.
The team was formed during the The Collider acceleration program and the Venture Builder program by the Mobile World Capital, which transforms scientific knowledge into cutting-edge technologies. According to CEO Miguel Silva, the team was formed when they realized the business potential for the technology they were developing.
"We discovered that a pharmaceutical company rejected 200 products in a day just because the packaging was damaged and it was nearly impossible to read the barcodes. Employees had to manually input that code, a job that doesn’t add any value,” said COO Adriaan Landman.
One solution, many sectors
Allread’s solutions can be used to track containers in ports, read utility meters with mobile phones, identify a product from a single photo, read damaged barcodes and even automatically spot medicines with a past expiry date.
The company’s algorithm was first tested in the energy sector to read the gas meters of Spanish natural gas and electrical energy utilities company, Naturgy, but more business opportunities quickly followed. After speaking with over 100 companies from diverse business sectors, the team considered the logistic sector as the one with the highest potential to be disrupted with its technology.
Last year, the Allread MLT started to test its solutions with the Port of Barcelona, where the greatest concern was pollution from freight transport.
“Our technology, was used to track each container departing the port [via railway], enabling information monitoring and data collection needed to reduce pollution” said Silva-Constenla. During the pilot, the Port of Barcelona’s estimated savings in the cost of identification and translation of alphanumeric text was about 80%.
In the warehousing sector, Allread MLT’s technology improved productivity in a Spanish warehouse located in the Girona region by 76% by decrypting damaged barcodes and automatically sending out orders to print new barcode labels. The same solution can also used in supermarkets to minimize delays caused by damaged barcodes of products.
Seeking funding for national expansion
With nearly every company looking for opportunities to digitize operations, especially in industries where efficient process management is crucial to boost productivity and customer satisfaction, Allread MLT is fast gaining traction in other industries.
The company was recently a finalist in the Hangar51 program launched by the International Airlines Group (IAG) in collaboration with Vueling and Iberia. Its technology has been implemented in IAG Cargo to identify and manage container inventory. The solution improved productivity by 75% over previous processes involving manual recording and uploading of container codes.
The company is now looking to open a new round of investment of about €500,000 to expand the technical team and undertake new projects at national level. "There is a lot of market potential in Spain. For such a young company it is dangerous to internationalize too quickly, first we want to work hand in hand with companies in the country” said Landman.
Silva-Constenla is convinced that the Allread MLT team’s cumulative experience, along with the support of a top-notch research center, will help the company to scale rapidly and adapt to different industrial scenarios even faster. “Until now computer vision solutions were limited to OCR and have been focused mainly on cameras and hardware components. We want to make a hardware agnostic system," the CEO said .
We pride ourselves on the accuracy of our information and reporting. Please help us by letting us know of any incomplete or inaccurate information on our website.