Proprietary Deep Learning Technology
Automated continuous learning process and improvement through the latest AI technology
The Moonvision software specializes in quality control and assurance of assembly products and surfaces.
All data (image and video) are collected and processed via a visual data lake using Deep Learning to ensure a continuous learning process and continuous improvement in quality control. Since our technology is coordinated with the latest AI technology, the potential of our software includes an exponentially increasing learning capability and enables our model training to be the best possible Few-Shot object recognition on the market with minimal effort and data.
In order to localize the product-specific features, a Labeling Tool is available which can be continuously adapted to each quality characteristic and which carries out a very precise and fully automated annotation of each deviation.
| Data Management | Explanation | USP |
|---|---|---|
| Team and Role Based Access Management | Data protection and processing can be assigned to specific units, persons and used across locations for fast information exchange and best information security. |  |
| Infinite Versioning | Our solution allows an error-tolerant generation of training data, since any data input can be revoked. | |
| Multiple Geometric Shapes | The Moonvision Toolbox enables the identification of all types of features with different geometric figures. This ensures that actual and target states can be recorded specifically and correctly for the application. | |
| Video Snapshot Extraction | Our company provides a tool, based on learning algorithms, which allows savings in hardware components, since our technology, can delimit processes in time and trigger snapshots at the right time, like sensors. | |
| Equivalent Inferences On-premise and Online | Our technology enables online and offline deploys with 100% equal results. You can perform online activities and virtual tests flexible and update the toolbox offline with the new results. With this we guarantee process reliability. | |
| Cloud Scale GPU Utilization | All data is collected and processed by a central cloud-based visual Data Lake. We provide analysis results quickly because GPUs are available permanently and scalingly. | |
| Machine Vision and IP Camera | We support GeniCam standards (e.g. Basler, IDS, ...) and common video streaming standards based on IP (e.g. H264, RTSP, HTTP). | |
| Transfer Learning | Our pre-defined software reduces the need of training data by a factor of 3-10, as we can apply industry-specific know-how for the specialization of models. This enables us to identify new and rare phenomena. | |
| Anomaly Segmentation | We can localize unknown and conspicuous features, such as conspicuous fiber structure, surface texture, component composition. | |
| Traceback Predictions to Single Training Samples | Our technology has automatic procedures for detecting incorrect entries in the training data. | |
| Model Introspection | The Moonvision software visualizes the functioning of image analysis steps, whereby model improvements are enabled systematically and not randomly. | |
Our benefits
Data Lake
Learn once, run everywhere
APIs
Support different ERP systems, automation systems and realtime messaging systems.
Individual
Custom configurations for your specific quality criteria
Allrounder
Support of most production processes
Data Reports
Image based quality assurance reports
It is important for our company that every company can use our software with simple resources.
This, as one of our goals, allows every camera system to be linked to our software and allows us to remain hardware independent. By choosing image processing systems freely, we can process features down to the microscopic level and identify every critical detail.
Since our software is self-sufficiently designed, it is the perfect solution for every company and every industry. The versatile application possibilities of our product allow our products to be integrated into any manufacturing process. Thus, our solution can be coupled simply and by small measures into a production process both at arrival and departure of goods as well as after each processing step. Due to the different industries, we also consider it necessary to analyze moving products in the production flow in addition to the standing product controls in order to avoid any disruptions.
Due to the great diversity of our technology, it is sustainably designed and can be optimised for operation depending on the requirements.
