Introduction
Data coming from sensors, images and sounds can be easily recorded and stored, but the crude information is useless in most cases if not adapted / transformed into valuable information. Sound or image for example have to be transformed and / or recoded before being used. This is generally known as “labelling”. Labelling is the identification of bio-responses (images, sounds...) referred to gold standard (behavior, bioparameters in blood samples...). Labelled data are the basis for the development of algorithms able to predict and manage animal health or welfare or take control actions (climate control, feeding strategies).
For example, using on-line pig sound analysis is a useful tool to predict possible lung infection and then, prevent use of antibiotics or reduce their use. A possible application is presented on the following figure:
