This collaborative project between Alliance R&D, IFIP and INRAE has developed an innovative solution for high-throughput phenotyping of sow farrowing behavior. The system consists of a box containing a Raspberry Pi microcomputer and two cameras, enabling two sows to be monitored simultaneously, day and night.
The collection and annotation of over 130,000 images from 10 farms has enabled the creation of a high-performance image analysis algorithm in terms of accuracy and analysis speed. This algorithm is capable of differentiating between eight postures (sitting, standing, kneeling, lying on the belly, lying on the right/left side with or without visible teats), with an accuracy of over 90% under a wide range of farm conditions.
Thanks to access to this new information, this project will contribute to a finer selection on maternal behavior, linked to piglet survival under the sow.
© NUCLÉUS 2018 - Legal mention - Creation by Mediapilote - Privacy policy - Cookies policy