Deep Learning Methods

Deep learning is a powerful technology that complements MVTec’s comprehensive machine vision toolkit. Customers can develop holistic applications using deep learning in combination with conventional machine vision.

Deep Learning Takes Data Classification to the Next Level

The machine vision software products from MVTec come with pre-trained CNNs (Convolutional Neural Networks), that allow applications to be developed with a relatively small number of training images. By using these networks, users can train their own classifier to classify new data. Provided with a sufficient amount of images, the training process automatically extracts each class’ distinct features. HALCON and MERLIC then analyze these images and automatically learn which features can be used to identify the given classes.

Advantages of Deep Learning

  • Automatic feature extraction
  • Reduced effort of programming
  • Big amounts of data can be used
  • Reduced development time

Deep Learning Methods in MVTec Products

The training for these deep learning methods can be performed on GPUs, as well as on CPUs. This greatly increases your flexibility in implementing deep learning, because training can also be performed directly on the production line. This makes it possible to adjust the application to changing conditions of the environment. The inference can be performed on GPUs, on x86 CPUs and on Arm® processors.

Thanks to our AI Accelerator Interface (AI²), we also support a growing number of AI hardware accelerators to increase the inference part of deep learning applications.

Why use MVTec Software and not Open Source?

  • Industry proven and copyright-free pretrained networks, optimized for typical industrial use-cases
  • Deep Learning Framework allows experienced users to create their own models
  • Free professional support from our experts worldwide
  • Maintenance and backward compatibility within different versions
  • Support of AI inference acceleration hardware