Monitoring pig production using learning-based machine vision
Updated: Jul 30, 2020
In a previous blog article it was described how High Performance Computing can be used on pork processing plants to gain knowledge about factors influencing carcass and meat quality and to facilitating the process of meat distribution. We moreover learned that artificial intelligence is a helpful tool in precision livestock farming (PLF) to analyse the large amount of data generated on pig farms to aid farmers in their decision making.(Read article here)
But how does this work? In the Cybele project, the goal is to work towards integrating machine learning techniques within PLF, to not only continuously monitor pigs during their growth but also to improve the early detection of diseases or behaviour changes. In the project, deep learning-based machine vision will be used on video and image data especially aiming at: monitoring the live weight of pigs, recognition of individual pigs and the early detecting of social harmful behaviours such as tail biting. Harmful behaviours such as tail biting are a major welfare and economic challenge in modern pig production. In barren environments, harmful behaviours often develop out of an enhanced motivation to explore, but can further progress into damaging behaviour directed towards pen-mates. Harmful behaviour does no only indicate an underlying welfare problem in groups of pigs, but the inflicted wounds are painful and can get infected. Monitoring individual pigs in order to timely detect health and welfare problems might enable farmers to take immediate management actions and with this, decrease health and welfare issues on-farm.