Diota accelerates the deployment of Augmented Reality solutions with the integration of a Deep Learning option allowing an automatic initialization of the tracking model that is more robust to changes in the environment
Initialization of the tracking model is a required step before starting to use augmented reality to display work instructions on the field. Different methods are available to perform this initialization. Standard learning techniques make it possible to set up a reference model proposing rapid and efficient initialization. During the deployment phases though, when systems must be operational in varying conditions, e.g. day, night, natural lighting, artificial lighting, or within numerous workstations with different assembly configurations and changing backgrounds, standard learning procedures must be enriched to take these variations into account.
To face these challenges, a novel groundbreaking approach has been integrated in our Diota 4X solution using deep learning technology to ensure easy initialization of tracking and robustness to changing environment.
The Deep Learning option offers an alternative to our standard learning approach offering even more flexibility during deployment of use cases in factories. Indeed, the Deep Learning approach allows to create a learning base from a small volume of data using few video recordings in combination with the digital mockup to offer a generic initialization method that adapts to a broad variety of variations. Thus, the customer is spending significantly less time during initialization and therefore saves time on deployments while having flexibility in the choice of workstations.
Deep learning option is also particularly relevant if you already have implemented our solution and you want to quickly replicate and deploy the same use case on another industrial site with a different work environment.