Robovision Deep Learning
Delivering added value from your data
What is Robovision?
The development of deep learning has also heralded the advent of vast datasets with which to start the process – and this data must be labelled and classified in order to progress. Robovision allows this to be automated - layers of artificial neurons operate in similar ways as in our biological brains and create the rules autonomously. In short, Robovision have developed a stack that makes deep learning labelling very user friendly, and allows you to manage thousands of labellers, operating on huge datasets.
Scan is able to offer this solution in conjunction with our guided proof-of-concept in order to provide excellence in hardware, using the likes of NVIDIA DGX-2 and ease of use in software to give your organisation a critical advantage in the journey to become a more AI enabled enterprise.
The Robovision platform is scalable in many aspects. The data curation phase on-boards your data on the fast flash-based storage systems of their storage partners (usually Pure Storage or NetApp), and then via a user friendly wizard you can activate a labelling session that gives you access to more than 60.000 labellers worldwide. While the labelling operation is ongoing, the platform scans for bias within the labelling crowd and uses deep learning in the background to generate a predictive model that proposes new labels in the next image, dramatically speeding up the session.
While the data is undergoing labelling, it is also being evaluated by labelling managers. Once approved, a cloud AI store is enabled and one can browse through different top notch deep learning models and evaluate how they perform on the data.