The BSc thesis was conducted under the supervision of Prof. Dr. Harald C. Gall and Dr. Pasquale Salza from s.e.a.l.
Janosch Baltensperger's work proposes a detailed workflow to bring deep learning models into production, as for the traditional software development. Deep learning has gained immense attraction with the emergence of big data and advanced computing power. But it has become clear that neural networks impose various additional challenges, and the complete end-to-end development process remains unspecified.
The work «Continuous Deep Learning. An in-depth investigation of the deep learning workflow» transferred the conceptual idea into practice by building a prototypic deep learning system, using some of the lastest technologies on the market. To examine the feasibility of the workflow, two use cases are applied to the prototype, based on a text classification and image processing problem. This work has the potential to be a guideline for the deep learning development process, for both research and industry practitioners.