Skip to main content
Conference abstract, 2016

Studying cow behaviour: use of convolutional neural networks and deep learning algorithms

Guzhva, Oleksiy; Herlin, Anders Henrik; Nilsson, Mikael; Ardö, Håkan; Åström, Karl; Bergsten, Christer


The combination of machine learning and pattern recognition applied to video data acquired from surveillance cameras is a promising concept for Precision Livestock Farming (PLF) research area. Continuous automatic recording of different social interactions and analysis of the behavioural data could save many men-hours and could be used for the monitoring and assessment of dairy farm performance as well as for the evaluation of animal health and welfare. This study is one of the steps aiming towards the development of a fully automatic system for studying dairy cow interactions. Video data obtained during our previous studies helped in creation of a database with video sequences needed for training the classifier and extending the range of parameters for monitoring purposes. In present study, social interactions were supervised in the waiting area before four automatic milking stations. Three top-down view cameras were used to collect video data. A seven-point shape model for each cow was developed to manually tag cows in video sequences. A classification system, investigating features from pairs of cow shape models, was also developed. The identification of social interactions, using the system and crosscheck with ethogram containing descriptions of all the interactions, was tested on a set of video sequences. The system showed the potential for further development and variability of results depending on a complexity of observed interactions. The proposed seven-point model may act as a crucial component towards building the robust system for automatic detection of social interactions in areas of interest.


Cattle behaviour, advanced mathematics

Published in

Publisher: DairyCare


DairyCare Working Group 2 Focused Meeting Activity Measurements in Ruminant Research and Beyond, 2016-06-20--2016-06-21, Leeuwarden, The Netherlands