Automatic cattle activity recognition on grazing systems

The use of collars, pedometers or activity tags is expensive to record cattle's behavior in short periods (e.g. 24h). Under this particular situation, the development low-cost and easy-to-use technologies relevant. Similar smartphone apps for human recognition, which analyzes data from embedded triaxial accelerometer sensors, we develop an Android app cattle. Four main steps were followed: a) acquisition model training, b) c) deploy, d) utilization. For acquisition, developed a system three components used: two smartphones Google Firebase account storage. generated database was used train recurrent neural network. performance training assessed by confusion matrix. all actual activities, trained provided high prediction (> 96 %). deploy using TensorFlow API. Finally, cell phones (LG gm730) test six Holstein cows (3 lactating 3 non-lactating). Direct non-systematic observations animals made contrast activities recorded device. Our results show consistency between direct our app.