Applied image recognition: guidelines for using deep learning models in practice

In recent years, novel deep learning techniques, greater data availability, and a significant growth in computing powers have enabled AI researchers to tackle problems that had remained unassailable for many years. Furthermore, the advent of comprehensive AI frameworks offers the unique opportunity for adopting these new tools in applied fields. Information systems research can play a vital role in bridging the gap to practice. To this end, we conceptualize guidelines for applied image recognition spanning task definition, neural net configuration and training procedures. We showcase our guidelines by means of a biomedical research project for image recognition