Technology: Training Tool
We do this by developing:
  • cloud-based training tools so anyone, using Web interface, can build own neural network or a trained algorithm
  • new, optimized formats for the storing of trained neural networks and machine learning algorithms
  • the storage for training datasets so they are easily accessed
  • import file adapters for popular deep learning frameworks
TRAINING TOOL
is a cloud-based storage for machine vision skills with convenient interfaces.
The primary purpose of the Cloud is to store the information/machine vision skills amassed through the training of neural networks. This allows customers to easily access and use the skills for their purposes.
The Cloud allows to:
  • teach neural networks to customer-specified scenes using Web tools with visualization features
  • save, update and add the created vision skills
  • upload the preferred vision skills to devices using customer delivery modules
  • adapt the vision skills base to other deep learning frameworks such as Caffe and Teano
  • exchange vision skills with other Cloud Hub customers
  • receive updates on the scenes trained by Scorch
  • send new/unknown scenes to the Cloud Hub so they are learned