Our proprietary Artificial Intelligence technology revolutionizes the automatic identification of objects and people within images.
DART enables direct access to images with the same content, ensuring maximum accuracy and enormously reducing the costs associated with the learning phase of Artificial Intelligence.
This allows to face problems that cannot be solved with traditional AI, considerably increasing the areas of application of vision systems.
Traditional AI systems need hundreds, thousands of matching sample images to train the system.
DART allows you to train with a few images (possibly only one) for each class.
Traditional computer vision involves monolithic training. To add a class you need to do all the training again.
DART offers on-the-fly training: you can dynamically add new classes without having to train the system from scratch.
Traditional AI provides classifiers, not identifiers.
As the number of classes increases (and the difference between them decreases), the number of false positives explodes.
DART is essentially an identifier.
The number of false positives is minimized and independent in the number of classes.
It is not possible to manage a large number of classes with traditional AI without incurring significant performance degradation.
DART is independent in the number of classes to be identified, providing the same high accuracy even with hundred thousands classes.
Traditional AI systems often provide incorrect identifications, especially if the classes to be identified are similar to each other.
DART is designed to provide only accurate and robust identifications. Regardless of the number of classes and their similarity.
Each computer vision solution presents unique challenges.
DART allows you to achieve the best results in terms of identification and response times by changing a few simple parameters.
Our solutions are Software as a Service (SaaS).
The images can be acquired from different systems such as lightbox, overhead cameras, smartphone app, ... and are sent to our CLOUD for processing.
The result of the identification or quality control is then returned to the user for viewing.
We also develop on-edge solutions for particular performance needs in terms of speed and accuracy such as product line control, sorting or quality control of digital prints.
Identification is only the first step of a complete and valuable solution. Custom quality control and compliance functions can be added. For example, product compliance control, reading of the status of electrical panels, and much more.
All our solutions are accessible through simple and intuitive Web Dashboards thanks to our APIs in REST standard.
The dashboards are fully customizable according to specific solutions, here some examples.