Competencies
Deep learning
Our platform - machine learning for industrial use
MABRI.VISION has developed the DEEP.MV software platform for the application of modern deep learning and AI technologies in the industrial environment. The platform combines all the necessary components and enables use at the plant level.
Possible uses
Our deep learning platform is ideal for applications where rule-based approaches are challenging or fail. This applies above all to test applications with a wide range of variants and error characteristics, such as varying surface and structural properties. The MABRI.VISION platform can be used optimally where, due to the complexity, primarily manual and manual checks are carried out using error catalogs.
Classification | Detection and classification of objects and errors
The classification forms the basis for many quality assurance tasks. In the application, there is no need for time-consuming labeling of data - it is sufficient to categorize images.
Error detection in flushing tabs
scratch
started
not OK
Web error
OK
Web error
Hole
Object detection | Examples: completeness check, attendance check, counting, pick & place and much more.
Completeness check
Position check
Attendance control
Defect detection | Surface defects, incomplete components, structural defects, color and shape defects and much more.
Injection molding defects
scratch
Pores and grooves
Segmentation | Detection of structures and image areas, surface defects, incomplete components, structural defects, color and shape defects and much more.
Segmentation is always important when objects, structures or errors need to be identified and isolated very precisely. This information can be used, for example, for process optimization
Chip detection
textiles
Rinse tabs
Incorrect sorting
foreign body
Anomaly detection | Errors on packaging, print image inspection, weaving errors and much more.
Circuit boards
Microstructures
Surfaces (e.g. adhesive residue)
OCR | Text recognition
DOT code on tires
Font recognition
Your advantages
Everything from a single source
MABRI.VISION provides you with everything you need for your application. In this way, you avoid using partial and isolated solutions that do not interlock and are therefore difficult to maintain. In addition to the software application, we primarily supply the necessary sensor systems, test platforms and services for your production process.
Save costs
By using our lean AI platform, challenging applications in the industrial environment can also be solved cost-effectively. Manual and time-consuming test steps can thus be omitted.
Solving complex problems
Deep learning technology can be used to solve applications that are very challenging or cannot be solved satisfactorily with classic image processing. This opens up completely new possibilities in quality assurance.
The right solution
As a universal solution provider for optical production metrology, we take a close look at your application. We only use our deep learning technologies when and to the extent that it makes sense for the application. In this way you avoid incorrect results and excessive development costs. In addition, we map all other essential components in quality assurance, including calibration and traceability.
Stay flexible
One challenge of the current development trends in the industry is modular production with ever shorter production cycles. Machine learning processes keep you flexible and allow you to adapt your production adaptively and independently.
Future proof
The development in the field of deep learning is progressing very quickly. Our AI platform is therefore constantly being expanded and relies on established standards in research and development. This continuous development enables you to always use the latest technologies in the future.
Practical example
Machine learning processes can be used wherever rule-based approaches are difficult to implement.
Error detection on socks
Web error
Hole
Defect detection on metal pipes
Span
Weld seam defect
Process flow
Deep learning methods are based on artificial neural networks. These networks are inspired by (biological) neural networks, i.e., part of a nervous system. Artificial neural networks are structured in layers. A layer, or level, consists of several artificial neurons. Between an input and output layer, there are several hidden layers. The name "deep" neural network derives from these hidden layers.
