Using vision-related behavioral traits for commercial breeding programs for pigs

Offentlig prosjektsammendrag

Poor economy and increased consumer demand for animal welfare friendly meat production has changed the framework for pig production. The farms are bigger with high animal turn-over. Manual work is automatized and the human-animal interactions are reduced. These factors are all creating the framework for the environment pigs are kept in. It is therefore crucial that breeding companies have information about behaviour to evolve the pig breeds to be calm, social and well-functioning in their social environment. This means to minimize antagonistic and damaging behaviours such as aggression and tail biting and stimulate to optimal behaviours such as play, positive interactions and exploration. Advances in vision technologies, artificial intelligence (AI) and machine learning (ML) have offered new opportunities to measure behavioural traits. In this project we aim to measure behavioural traits in pigs that influence animal welfare, production efficiency and product quality in order to establish a routine pipeline for implementation of behavioural traits in our breeding program. Initially, the project shall gather data from Norsvins boar testing station through cameras installed over each pen of spesific sections of the test station. At the same time, the project will begin the building of a system based on machine learning for efficient selection and storage of relevant images of behavior of pigs. Further, the videos will be manually annotated by researchers for creating a powerful training dataset for the algorithms. The videos will also be used for identification of new traits that can be relevant for a future breeding goal.