Work

Accelerating Utilization of Computer Vision and Value Creation

Forestry and Biomaterials Industry

Process and production monitoring at factories is typically done manually, and 24/7 attention is required by the control room personnel to carry out this task. The operators are overwhelmed by a high number of displays and the information they provide, and their mobility is restricted.

The Developed Computer Vision Platform enables the customer to scale computer vision utilization across the company operations. Users are provided tools to easily deploy or fine-tune existing models as well as develop and onboard their own models.

Read the full reference case here.

Automated Quality Assurance Process

Automotive Product Manufacturing

Quality assurance in automotive product manufacturing requires adaptation to a plethora of components and parts as well as tight quality requirements. This combination makes manual inspections time-consuming and rule-based automation difficult to implement.

The developed computer vision solution enabled the customer to roll out an automated quality assurance process across their sites globally. The technology approach combines deep-learning-based computer vision with an easy-to-use cloud-based software application.

Read the full reference case here.

Minimizing errors in PCB assembly

Electronics Manufacturing

Through-hole technology (THT) is widely used in the assembly of printed circuit boards (PCBs) for products with high power and physical robustness requirements and often relatively low production volumes. Any defects in the component assembly process can prove costly by leading to reassembly steps or even defective end products. Furthermore, new products are regularly introduced in production, which would make automated systems without lifecycle considerations quickly obsolete.

In addition to automating the visual quality inspection of a THT assembly with computer vision technology, Top Data Science has helped customers to introduce a streamlined practice for introducing new products in the process. Besides directly reducing the cost of quality, this supports the core business need of a dynamic production portfolio.

Improving construction productivity

Construction industry

Seismic reinforcement of concrete bridges involves attaching metal brackets to enhance the stability of the structure during seismic activity. Traditionally the exact locations of the support brackets are measured using film and manual markings, a laborious process that is prone to errors.

To address these challenges for a Japanese engineering leader, Top Data Science developed a solution that automates the measurement process using smartphone camera images. With this approach, a construction worker can obtain the measurement results simply by taking pictures with a smartphone. The results are calculated in seconds, drastically reducing time and minimizing errors.

Read the full reference case here.

Ensuring quality in logistics

Chemical industry

Efficient logistics of manufactured goods are often reliant on printed package labeling. Even minor glitches in the physical package labeling process can result in the product ending up with the wrong recipient, perhaps in a geographical location far from the correct one. These events, even when happening rarely, can have a significant negative impact on customer satisfaction and be associated with a meaningful extra cost.

To address these challenges for a Finnish food ingredient manufacturer, Top Data Science developed a computer vision -based solution that automates the visual checks of package label content and readability. Comparing critical information detected in the label with the ground truth of the production system, it can immediately alert the production operators about any issues or even control the production line directly.

Pushing technological boundaries in traffic safety

European Research

While the grand goal of driving technology development is a fully self-driving car, AI is enabling significant safety improvements also on the lower ladders of autonomy. In the EU-funded Shape Future project Top Data Science has developed computer vision -based technologies for a safer driving experience.

The main focus of the work has been automated traffic blind spot indication with depth-based video analysis. Top Data science has also contributed to research on virtual blind spot removal, a cloud-based capability that utilizes external video sources to analyze the risk levels associated with different blind spot scenarios.

Optimized Pulp Production Process

Forestry and Biomaterials Industry

Pulp production process requires 24/7 control to ensure that the end-product quality meets the requirements of customers with high certainty. At the same time the use of chemicals and other raw materials should be minimized to support achieving financial and sustainability goals.

Developed deep-learning based predictive modelling application framework enables optimization both for pulp and board production, and it can be utilized in other domains due to its modularity and customizability. Optimization services can be used through an easy-to-use end-user application or as an embedded service.

Automated Meter Reading Service

Energy Utilities

A typical power plant has a plethora of analog and digital meters whose values have to be read on a regular basis with adequate accuracy. This is a repetitive and time consuming activity. The meters measure many different things from temperature to moisture, pressure and electricity consumption.

Due to the limited amount of data we decided to use multiple computer vision methods, which enabled us to achieve accurate and reliable results also on edge cases. The methods included Feature Matching to extract the Region of Interest (RoI) precisely, and a generative model approach for edge cases.

Read the full reference case here.