Beef Cow Behaviour Prediction using Deep Learning

Overview

This research explores the use of highly granular accelerometer data to predict and classify specific behaviors in beef cows within both commercial and research settings.

Technical Approach

• Signal Processing: Application of Low Pass Filters and aggregation techniques to clean noisy sensor data.
• Time Series Modeling: Comparing the performance of various deep learning architectures:
  • RNN (Recurrent Neural Networks)
  • LSTM (Long Short-Term Memory)
  • BiLSTM (Bidirectional LSTM)
  • Transformers
• Objective: To determine the most effective model for behavior exploration and utilization in precision livestock farming.

Results

The study provides a comprehensive evaluation of how different deep learning techniques handle high-frequency time-series data for animal behavior monitoring.