Deutsche Telekom last week announced a new service that provides AI developers with the ModelArts platform from the Open Telekom Cloud.
In the beta version, users can test it free of charge. Access is via the Open Telekom Cloud website, or now directly via the new AI homepage.
With the auto-learning function, developers can train an AI model even without much experience. Different use cases can be activated with one mouse click. Object and image recognition are available as well as predictive analytics. These are predictive analyses that are used, for example, to calculate maintenance intervals. Only three steps are necessary until the AI recognizes objects. In the first stage, the program receives several images with the marked object. In stage two the algorithm trains the object recognition. And in the third stage the user generates his program and can test its learned "abilities" on unknown images.
Professionals can also access pre-installations of common open source frameworks (utility programs with open source code). Exemplary are mentioned here: Tensorflow, PyTorch, Mindspore, MXNet and Caffe. They can immediately work with their own code. In addition, ready-made algorithms, which only need to be trained accordingly, shorten programming times. With virtual OpenSource Jupyter notebooks, they have access to the required computing and storage power.
The preparation of the data and the training of the AI takes place exclusively in the Open Telekom Cloud.