Tensr Consulting

tensr — Fri, 28 Nov 2025 08:23:15 +0000

From Pixels to Parking

A Smarter Approach to Parking Enforcement in Brussels

Finding a parking spot in Brussels can be a real challenge. And for Parking Brussels, the organization responsible for managing on-street parking, it’s not just about finding a spot – it’s about ensuring fairness, efficiency, and accurate enforcement. With millions of parking events each year, they knew there had to be a smarter way to manage things. This is where the power of new technology came in, with a little help from Tensr and Google.

The Customer

Parking Brussels is responsible for managing all on-street public parking in the city. Their role includes ensuring that parking spaces are available and used correctly, and that any violations are handled appropriately. This is a big job that involves a huge amount of data and a constant need to improve accuracy

The Problem:
More Data, More Complexity

Parking Brussels uses “scanning vehicles”, vehicles equipped with cameras that capture images of parked cars. This data is then reviewed to ensure parking regulations are respected. The agency sought to empower its human reviewers with more insightful information to help them assess each situation accurately and efficiently. The goal was to provide tools that could help reviewers better interpret complex scenarios, such as nuanced license plate details, vehicles engaged in permitted activities like loading/unloading, or the presence of valid disabled parking permits

The Goal:
Partnership and Purpose-Built AI

Parking Brussels knew it wanted to explore AI but needed help understanding its potential and how to apply it to their specific challenges. Because of our expertise in building practical, results-driven solutions and our strong partnership with Google we were the perfect partner for this challenge.

Four key use cases

Our collaboration focused on how AI could enhance the ability of human reviewers to interpret complex parking scenarios, specifically in four key areas:

Enhanced detail extraction for vehicle identification

Improving the clarity and accuracy of crucial vehicle identifiers, like license plates.

Contextual Scene Analysis

Detecting relevant activities or objects in the immediate vicinity of a parked vehicle, providing reviewers with a more complete picture of the parking situation.

Vehicle Interior Assessment Support

Assisting reviewers by providing information about the potential presence of occupants within a vehicle.

Detection of Official Authorizations

Improving the system’s ability to identify official permits or placards displayed on or within vehicles.

The project leveraged various Google technologies, including Gemini for its capacity for detail recognition on partially obscured images, Vertex AI AutoML to train custom models, and the Vision API to detect broader aspects of the images.

The Solution:
Tangible Benefits and a Clear Path Forward

The solution demonstrated significant potential for improving the accuracy and efficiency of the review process. The AI-assisted approach showed an improved ability to present clear license plate information, aiding reviewers in their assessments. The project reaffirmed that the technology is designed to support, not replace, human oversight.

The success of the project validated how the use of AI could directly lead to:

Improved citizen experience:
By equipping reviewers with more comprehensive information, the system aims to support more accurate fine assessments, leading to less hassle for drivers regarding potentially incorrect tickets and fewer administrative follow-ups for Parking Brussels.
Fairer revenue collection:
Enhanced accuracy helps ensure parking regulations are applied correctly and revenue is collected fairly.
More efficient operations:
The project indicated potential for significant time savings and operational efficiencies, allowing staff to focus on more complex cases and streamlines the overall enforcement support process.

A Look Ahead

The Parking Brussels project is a great example of how AI can be used to solve real-world problems and deliver tangible benefits. It’s not about replacing humans – it’s about equipping them with better tools and insights. As AI continues to advance, we expect to see even more innovative applications that make infrastructure smarter, more efficient, and more citizen-friendly.

Ready to unlock the power of visual data for your organization? We specialize in crafting tailored solutions that solve real business challenges and drive meaningful results. Our solutions are designed with data privacy as a top priority, ensuring your sensitive data remains secure and fully compliant with regulations. We would like to thank GC Innovate for facilitating this great collaboration.

The post appeared first on Tensr Consulting.

tensr — Fri, 17 Oct 2025 11:19:50 +0000

Beyond the Naked Eye

How Visual AI Solves Real-World Problems

We live in a world of data. While we’re used to seeing data in spreadsheets and structured texts, the biggest untapped source of information is all around us, often invisible to the naked eye. We’re talking about visual data. Imagine the millions of images captured daily by cameras and sensors, or the countless handwritten documents sitting in archives. At Tensr, we believe that unlocking this visual goldmine is the key to tackling complex social and business challenges. With our deep expertise in image analysis and advanced AI systems, we transform raw visual information into valuable, actionable insights.

The Unseen Problem: Where Traditional Methods Fall Short

Traditional processes are often manual, time-consuming, and prone to error. They simply cannot handle the sheer volume of visual data being generated every day. For instance, consider the challenge of monitoring vast natural reserves for litter and pollution. An environmental organization faces the immense task of quickly and effectively mapping and analyzing huge areas—a job that’s nearly impossible to do by hand. This is a very real problem that shows how traditional methods are no longer enough. There’s a clear need for a smarter, more scalable approach to tackle these kinds of challenges.

A New Vision: Transforming Data into Impact

This is where our expertise in visual AI plays a crucial role. We build systems that can not only “see” images, but “understand” them. By leveraging cutting-edge technologies, we turn visual data into concrete, solution-oriented actions that complement, not replace, human efforts.

Let’s go back to the environmental organization. Instead of manually searching vast areas, a visual AI solution can analyze drone footage to identify the type and concentration of litter. The system can distinguish plastic bottles from metal or paper and create a detailed map of the contamination. This allows cleanup crews to plan their operations more efficiently, leading to a faster and more effective approach.

This is just one example. Visual AI can also be used to help farmers detect diseases on their crops early, or to assist the visually impaired in navigating their surroundings by recognizing objects and signs. The true power of this technology lies in its ability to make complex problems solvable.

Inside the Toolkit: The Power of Google’s AI

Our solutions are built using the most powerful tools on the market. We utilize advanced technologies, such as Google’s AI models, including:

Cloud Vision API: A ready-to-use, pre-trained tool that can detect faces, landmarks, logos, and text in images. It’s perfect for fast and cost-effective solutions.
Document AI: Specifically designed for document processing, this tool combines OCR (Optical Character Recognition) with machine learning to understand document structure and efficiently extract information.
Video Intelligence API: A powerful tool that analyzes video content in real-time to detect objects, activities, and scenes, ideal for applications ranging from media management to security.
Vertex AI: With this platform, we can build and train custom, deep machine learning models tailored to specific, complex challenges.
Gemini: For its unparalleled ability to provide detail and recognition, even on partially obscured or complex images.

These tools, combined with our expertise, allow us to build scalable, accurate, and robust solutions that turn the promise of AI into a practical reality.

The Road Ahead: Building a Smarter Future

The possibilities are endless. From analyzing video feeds to enhance safety in factories to automating document processing, visual AI has the potential to create a tremendous positive impact on our society. The core of this technology isn’t just about efficiency; it’s about creating a smarter, fairer, and safer world for everyone.

We believe that AI is a powerful tool, not to replace humans, but to empower them with better insights and tools. At Tensr, we are ready to build that future.

Curious how the untapped visual data in your organization can be turned into tangible value?

The post appeared first on Tensr Consulting.

tensr — Fri, 03 Jan 2025 10:21:28 +0000

Future-Proofing Data Processing

Enabling parallel processing and AI integration with Google Cloud.

The Customer

Spectricity is a fabless company that develops and manufactures spectral sensing solutions using CMOS technologies. Their solutions are designed for high-volume production and use in consumer and mobile devices. The company was incorporated in 2018, and is backed by venture capital. Its work builds on an exclusively-licensed patent portfolio from imec, a world-leading R&D and innovation hub in nano-electronics and digital technologies.

The Problem

Spectricity’s existing solutions were built with traditional data science tools on local devices, which limited their ability to fully realize the potential of their spectral sensing solutions. The large files they worked with could not be efficiently processed, and the main challenge was enabling parallel preprocessing of data in a way that integrated seamlessly with the rest of the pipeline.

The Goal

To design a future-proof framework that would allow Spectricity to build and deploy scalable, efficient ML solutions capable of handling large datasets, while improving the flexibility and efficiency of their data preprocessing processes.

Google Technologies used

AI Platform

Cloud Dataflow

Cloud Storage

The Solution

Tensr worked closely with Spectricity to capture their requirements through workshops and design a robust, scalable framework for ML solutions. The solution involved migrating data preprocessing to Apache Beam running on Google’s Dataflow service, enabling efficient processing of large HDF5 files. The files were stored securely in Google’s GCS data lake, while data for analysis and dashboarding was stored in a serverless DWH (BigQuery). This solution ensured parallel processing and improved data handling for AI models.

Customer gain

Spectricity’s pipelines evolved from small-scale, local ML processes to robust, distributed systems. These new pipelines are now scalable, easy to retrain, monitor, and manage, making them suitable for industrial-scale applications and allowing Spectricity to fully leverage the potential of their technology.

The post appeared first on Tensr Consulting.

tensr — Fri, 03 Jan 2025 10:20:35 +0000

Smart document processing

AI-Powered Document Processing for Smarter Health Insurance Workflows.

The Customer

The LCM is among the largest health care organizations in Belgium that offers quality health at a low cost, but also takes good care of its members. LCM works to assure improvements in well-being through the range of services for health and prevention, as well as financial assistance to maintain access to quality care for individuals and families in Belgium.

The Problem

CM processes 5 million handwritten reimbursement documents each month, but the current scanner often fails to recognize handwriting accurately. As a result, employees still spend a lot of time manually reading and typing information from paper documents into the system.

The Goal

Implement a reliable solution that accurately recognizes and processes handwritten reimbursement documents, reducing the need for manual data entry and enabling employees to focus on higher-value tasks.

The Steps

Discover & Inspire

Tensr organised workshops with two main goals:
understanding existing business processes and educating LCM on what AI is and how to apply it to gain business value from it.

Strategise & Ideate

Tensr was responsible for translating business requirements into functional and technical solutions. A data and AI strategy were developed to ensure long term success.

Prototype

To ensure success, various prototypes were developed. Initially these prototypes used existing APIs for rapid iterations. These benchmarks were compared to existing solutions to illustrate the added value and demonstrated superior performance of later versions. These prototypes were crucial to ensure feasibility of the solutions.

Build

Our consultants integrated with the LCM development team to build, train and deploy AI models in a largely automated way. Microservices for image processing were also developed, tested and released to support the AI solutions.

Deploy

As part of integrating with the LCM technology stack and to guarantee consistent performance and reliability, we defined and setup AI model performance tests. This ensures that the solutions are able to process up to 100,000 certificates per day, while maintaining visibility on performance, resource management and security.

Support & Train

To ensure the customer remains in full control, we provided monitoring and quality assurance tools. We also gave knowledge transfers as part of the handover process and still provide support for the customers production teams.

The Solution

Using advanced preprocessing techniques we extract handwriting from the documents. Each document passes through several smart models that categorise each document after which we extract relevant fields from each document and pass it through our deep ML models. Finally we determine if a document can be processed automatically or first needs to be validated by an operator.

Customer gain

Due to the addition of AI we’ve been able to speed up the intake process of health insurance documents with a higher percentage of forms recognized thanks to the AI. This eventually leads to a faster reimbursement of healthcare costs for the customers and higher customer satisfaction.

The post appeared first on Tensr Consulting.

tensr — Fri, 24 May 2024 13:27:19 +0000

Unified Architectural Data Aggregator

Streamlining data integration across varies sources and structures.

The Customer

UZ Leuven is an academic hospital in Leuven, Belgium, associated with KU Leuven. It consists of three campuses (as of 2022), providing approximately 2000 beds and enrolling 9000 employees. It is part of the Flemish Hospital Network KU Leuven, which is a partnership of 30 Flemish hospitals. KU Leuven is the biggest hospital in that partnership.

The Problem

Fragmented Data Sources: UZ Leuven faces the challenge of managing clinical data dispersed across multiple sources such as KWS (Sybase), LWS (Sybase), FHIR Datastores, Peoplesoft, eCFR and Formaza Documents (XML), leading to inefficiencies in data access and utilization.

The Goal

To establish a Centralized Data Warehouse UZ Leuven aims to develop a future-proof, scalable and secure Data Warehouse solution to consolidate clinical data from different sources. This initiative seeks to streamline data management processes and ensure governed access for researchers, doctors and clinical support managers (CSMs), fostering efficient and effective data-driven decision-making within the organization.

The Steps

Discover & Inspire

Tensr organised workshops to discover existing challenges and understand the as-is situation. Important aspects that were focused on include:

Current architecture
Existing tools and technologies
Existing technical expertise
Functional and technical requirements (legal, performance, standards, …)

Strategise & Ideate

Through interactive workshops we collaborated with the customer to build a high level architecture that could be reviewed and implemented by relevant teams such as infrastructure, networking and security.

Further workshops with more specialised teams served to ideate about how the data platform could be used to replace or improve on existing tools and processes. Through brainstorming sessions and workshops, we were able to point towards existing tools inside or outside GCP to solve a number of challenges (intelligent anonymisation, best practices on observability, issue management, data cataloging, etc).

Prototype

To illustrate some concepts, we provided prototype samples which could be used to discuss internally, certain concepts and investigate how they could be used as more efficient or robust solutions to existing scripts.

Build

The customer prefered to build and deploy the solutions themselves. Tensr is ready to assist as both a support or an implementation partner if desired by the customer.

Deploy

Deploying the solution and integrated it with the customer’s existing systems.

Support & Train

Provide ongoing support and maintenance for the solution and develop standard operating procedures and training materials.

The Solution

As a result, Tensr delivered a detailed description of the suggested future data platform from an architectural, functional and security perspective. Subsequent workshops provided the customer with the necessary tools to take full ownership of the project, taking into account Google best practices, industry-specific requirements and the highly sensitive nature of the data.

The Architecture

Customer gain

The implemented solution instils confidence in scalability and compliance is instilled within the customer. National and industry-specific policies are seamlessly adhered to. Furthermore, the platform actively enhances data quality for various users by introducing a standardized model. While not hindering the autonomy of business users who can still effectively do what they do best.

The post appeared first on Tensr Consulting.

LCM – The intelligent scanner

tensr — Wed, 15 Nov 2023 14:14:57 +0000

Even in 2023, the Christian Mutuality (LCM) receives about 40,000 mostly handwritten doctor’s certificates per day. These certificates contain information regarding the patient, the reason(s) for the doctor’s visit, and the amount to be paid for the doctor’s consult. Part of the amount is refunded by the LCM. The data of the doctor’s certificates must be digitized before they can be processed. For handwritten certificates, this requires manual labour to read from the certificate and input the data in the applications and databases. Automating the process of reading and transcribing the handwritten text reduces labour cost and speeds up the processing of the certificates.

An intelligent solution

The solution is based on artificial intelligence (AI), augmented with advanced image processing. It is capable of automatically reading and transcribing close to 20% of all A-type doctor’s certificates.
This results in a serious manual labour reduction and a significant shortening of the processing time, without loss of quality. Speedier refunds thus result in happier customers.
The AI is tuned in such a way that it does not make more mistakes than humans do, and it is monitored over time to ensure its quality does not deteriorate. Every now and then the AI needs to retrain to maintain its high standard and to handle previously unseen data.

“Thanks to the handwriting AI, we are now capable of automatically processing nearly 20% of all A-type doctor’s certificates, which greatly enhances and speeds up our patient reimbursement process.”

Roel De Spiegeleer, ICT director LCM

The “Landsbond der Christelijke Mutualiteiten” (LCM), or Christian Mutuality (CM) for short, is a Belgian health insurance fund that groups nineteen regional health insurance funds. With its four and a half million members, CM is the largest health insurance fund in the country. Its mission is “keeping healthcare affordable for everyone”.

More info on cm.be

2-step approach

Reading handwriting on forms requires a two-step approach:

1. Finding the text

Finding the text uses advanced image processing techniques. Doctors don’t tend to write between the lines, and they don’t use the same writer’s style. Therefore, a set of algorithms is used to detect text boundaries and to homogenize the text’s look and feel.

The image-processing algorithms are also required to extract a good training set for the AI models. Examples of such algorithms are:

Fetching the boundary box of a region of interest (ROI)
Detecting the number of text lines within the ROI
The splitting of text into words
Creating uniform pen thickness of the text
Suppression of background noise
Reduction of colours
Matching against templates
Alignments

2. Reading the text

Once found on the image, the reading of the text is performed by custom-trained AI, based on neural networks and optimized for handwriting. The architecture of the neural network combines recurrent and convolutional layers.

The custom-trained AI performs much better than available third-party generalized models, because they are trained for the specific business case and know the business jargon very well.

Both the image processing, which is CPU-intensive, and the transcribing of the handwriting, which is RAM-intensive, are offered as configurable, stateless micro-services accessible through a RESTfull API. They are integrated with a front-end and back-end application, and orchestration through an external workflow engine.

The production system is run completely in the local data centre. The AI model trainings are performed using Vertex AI on the Google Cloud Platform.

Applied Technologies

The microservices are written in Python and run using Flask servers within Docker containers in an OpenShift (Kubernetes) environment. This ensures horizontal scaling and optimized tuning for either the image processing towards CPU, or transcriptions towards RAM.

The image processing algorithms are built on standard Python libraries, such as OpenCV. The AI is built using TensorFlow and is trained using Google Vertex AI on the Google Cloud Platform. The AI monitoring and quality assurance is executed on a Looker Studio dashboard on the Google Cloud Platform.

Additional technologies are used for small dedicated tasks that support the AI solution. These include:

Tesseract OCR for the recognition of non-handwritten text;
Google Vision API for the recognition of handwritten text for which no custom model can be trained and for which no self-learning AI can be built;
YOLO for finding non-text objects on a form, such as stamps or signatures.

All code is integrated with the in-house software release management based on GitLab, Jenkins and Nexus.

A team within a team

LCM maintains a framework contract with Cronos Public Services, the sub-domain of the Cronos Group that bundles and facilitates the expertise of its sub-companies for public services. Thanks to Cronos and Vectr.Consulting, the Tensr team was integrated in the LCM development teams with the task of introducing everything AI-related that was needed for reading the doctor’s handwriting, including the continuous AI monitoring and learning capabilities. This was the largest project within the LCM to integrate such an AI within its back-end infrastructure.

The post LCM – The intelligent scanner appeared first on Tensr Consulting.

Lisa just joined our team as a jobstudent. Get to know her!

furora — Wed, 26 May 2021 09:25:18 +0000

We’re very excited to welcome Lisa Van Vlem, she’s our first job student at Vectr. ????
She’s currently finishing a postgraduate in #AI and coding at Erasmushogeschool Brussel. We’re looking forward to working together and learn from each other.

????️ Get to know Lisa in the video below, sound on! In the coming months she will take you along on her journey here at Vectr, so stay tuned and join us in welcoming her.

The post Lisa just joined our team as a jobstudent. Get to know her! appeared first on Tensr Consulting.

Advanced analytics in real estate

furora — Tue, 02 Mar 2021 16:02:54 +0000

Professional customers provide real estate listings for publication to Immovlan. Some listings appear to be a lot more popular than others in the searches of house seekers. We want to understand why this is in order to define guidelines for professionals customers to improve their chances of selling a listing

Solution: Listing optimization

By performing advanced analysis on visitor behaviour and listing properties we defined a set of rules to improve the chances of conversion. Using these rules we showed that by optimizing the details for each listing with regards to the type of listing, we could significantly improve conversion rate.

Valuable insights can be provided for paying customers.
Higher conversion rate for better platform performance.
Serverless Data Warehouse enables advanced analysis on existing but unused datasets.

Google technology

Big Query

Pub Sub

App Engine

The post Advanced analytics in real estate appeared first on Tensr Consulting.

Smart document processing

furora — Tue, 02 Mar 2021 16:02:11 +0000

Handwriting reimbursement documents need to be processed by CM in order to make paybacks possible. The current scanner solution of the CM processes 5 million handwritten documents per month but doesn’t achieve the requirements due to not always recognising the handwritings correctly. Consequently, employees still invest a huge amount of their time in deciphering and typing the text of the paper documents into their application.

Solution: Intelligent scanner with an AI add-on

By adding AI to the scanner it now becomes intelligent and self learning. It recognises what is written and makes suggestions on the screen of the employee before it automatically inserts it in the system.

Faster reimbursement of health care costs for people.
Speeding up the intake process of health insurance documents.
Higher percentage of forms recognized thanks to the AI.

“Our people process 5 million handwritten documents per month.
Thanks to adding AI to this process, our scanner recognizes a lot more!”
Peter Van Maele – Head of acrhitecture ICT CM

Google technology

Vision API

AI Platform

Big Query

Cloud Build

The post Smart document processing appeared first on Tensr Consulting.

Hadoop migration

furora — Tue, 02 Mar 2021 16:00:26 +0000

The amount of data grows exponentially (50% p.a.) which makes it very challenging and costly to maintain an on-prem data center. The existing infrastructure has once again reached its limits which forces RIPE NCC to look for better solutions for their numerous workloads.

Solution: Cloud Migration: Hadoop to GCP

By migrating the entire data workload to GCP, RIPE NCC gains peace of mind while their data grows exponentially. By separating the data storage from the computations, the migration to GCP also leads to a highly efficient cost model.

Better scalability in the face of exponential data growth.
Serverless tooling for maximum efficiency in day-to-day workflows.
Managed Services for hosting datasets to the public.

“We chose to rely on the Google Cloud Platform for all our data-centric workloads since we feel they offer the best value”

Google technology

Big Query

Bigtable

Dataflow

Kubernetes Engine

The post Hadoop migration appeared first on Tensr Consulting.