Prototyping a mobile app which detects dogs’ emotions based on their body posture: a design science approach

The automated analysis of animal emotions is of interest for a variety of purposes. It may be necessary, for example, for animal therapy to be able to classify and categorize emotions as they arise. Therefore a fast mobile solution is required. Hence, this chapter is about detecting the emotions of dogs based on their body posture. Additionally, a reliably labelled training dataset is required. We use and extend an existing dataset of dog images in the emotional states Anger, Fear, Happiness, and Relaxation. Additional images are collected in a crowdsourcing approach by joining dog-friendly social networks (online and in real life) to benefit on the one hand from their pictures and on the other hand from their collective intelligence. This dataset served as input for various machine learning models- commercial ones and a self-built model- used to predict the aforementioned emotions of dogs. In doing so, this chapter follows the design science research process resulting in an implementation of the app.