Deep learning based classification of sheep behaviour from accelerometer data with imbalance
Classification of sheep behaviour from a sequence of tri-axial accelerometer data has the potential to enhance sheep management. Sheep behaviour is inherently imbalanced (e.g., more ruminating than walking) resulting in underperforming classification for the minority activities which hold importance...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-09-01
|
| Series: | Information Processing in Agriculture |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2214317322000415 |
| _version_ | 1827810072165613568 |
|---|---|
| author | Kirk E. Turner Andrew Thompson Ian Harris Mark Ferguson Ferdous Sohel |
| author_facet | Kirk E. Turner Andrew Thompson Ian Harris Mark Ferguson Ferdous Sohel |
| author_sort | Kirk E. Turner |
| collection | DOAJ |
| description | Classification of sheep behaviour from a sequence of tri-axial accelerometer data has the potential to enhance sheep management. Sheep behaviour is inherently imbalanced (e.g., more ruminating than walking) resulting in underperforming classification for the minority activities which hold importance. Existing works have not addressed class imbalance and use traditional machine learning techniques, e.g., Random Forest (RF). We investigated Deep Learning (DL) models, namely, Long Short Term Memory (LSTM) and Bidirectional LSTM (BLSTM), appropriate for sequential data, from imbalanced data. Two data sets were collected in normal grazing conditions using jaw-mounted and ear-mounted sensors. Novel to this study, alongside typical single classes, e.g., walking, depending on the behaviours, data samples were labelled with compound classes, e.g., walking_grazing. The number of steps a sheep performed in the observed 10 s time window was also recorded and incorporated in the models. We designed several multi-class classification studies with imbalance being addressed using synthetic data. DL models achieved superior performance to traditional ML models, especially with augmented data (e.g., 4-Class + Steps: LSTM 88.0%, RF 82.5%). DL methods showed superior generalisability on unseen sheep (i.e., F1-score: BLSTM 0.84, LSTM 0.83, RF 0.65). LSTM, BLSTM and RF achieved sub-millisecond average inference time, making them suitable for real-time applications. The results demonstrate the effectiveness of DL models for sheep behaviour classification in grazing conditions. The results also demonstrate the DL techniques can generalise across different sheep. The study presents a strong foundation of the development of such models for real-time animal monitoring. |
| first_indexed | 2024-03-11T22:48:34Z |
| format | Article |
| id | doaj.art-435f6ee3b9b2413ea49006f355518103 |
| institution | Directory Open Access Journal |
| issn | 2214-3173 |
| language | English |
| last_indexed | 2024-03-11T22:48:34Z |
| publishDate | 2023-09-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Information Processing in Agriculture |
| spelling | doaj.art-435f6ee3b9b2413ea49006f3555181032023-09-22T04:38:47ZengElsevierInformation Processing in Agriculture2214-31732023-09-01103377390Deep learning based classification of sheep behaviour from accelerometer data with imbalanceKirk E. Turner0Andrew Thompson1Ian Harris2Mark Ferguson3Ferdous Sohel4Discipline of Information Technology, Murdoch University, Murdoch, WA 6150, Australia; Centre for Crop and Food Innovation, Food Futures Institute, Murdoch University, Murdoch, WA 6150, AustraliaCollege of Science, Health, Engineering and Education, Murdoch University, Murdoch, WA 6150, AustralianeXtgen Agri, Saint Martins, Christchurch 8022, New ZealandneXtgen Agri, Saint Martins, Christchurch 8022, New ZealandDiscipline of Information Technology, Murdoch University, Murdoch, WA 6150, Australia; Centre for Crop and Food Innovation, Food Futures Institute, Murdoch University, Murdoch, WA 6150, Australia; Corresponding author at: Discipline of Information Technology, Murdoch University, 90 South Street, Murdoch, WA 6150, Australia.Classification of sheep behaviour from a sequence of tri-axial accelerometer data has the potential to enhance sheep management. Sheep behaviour is inherently imbalanced (e.g., more ruminating than walking) resulting in underperforming classification for the minority activities which hold importance. Existing works have not addressed class imbalance and use traditional machine learning techniques, e.g., Random Forest (RF). We investigated Deep Learning (DL) models, namely, Long Short Term Memory (LSTM) and Bidirectional LSTM (BLSTM), appropriate for sequential data, from imbalanced data. Two data sets were collected in normal grazing conditions using jaw-mounted and ear-mounted sensors. Novel to this study, alongside typical single classes, e.g., walking, depending on the behaviours, data samples were labelled with compound classes, e.g., walking_grazing. The number of steps a sheep performed in the observed 10 s time window was also recorded and incorporated in the models. We designed several multi-class classification studies with imbalance being addressed using synthetic data. DL models achieved superior performance to traditional ML models, especially with augmented data (e.g., 4-Class + Steps: LSTM 88.0%, RF 82.5%). DL methods showed superior generalisability on unseen sheep (i.e., F1-score: BLSTM 0.84, LSTM 0.83, RF 0.65). LSTM, BLSTM and RF achieved sub-millisecond average inference time, making them suitable for real-time applications. The results demonstrate the effectiveness of DL models for sheep behaviour classification in grazing conditions. The results also demonstrate the DL techniques can generalise across different sheep. The study presents a strong foundation of the development of such models for real-time animal monitoring.http://www.sciencedirect.com/science/article/pii/S2214317322000415Sheep behaviour classificationData synthesisClass imbalanceGrazing sheep |
| spellingShingle | Kirk E. Turner Andrew Thompson Ian Harris Mark Ferguson Ferdous Sohel Deep learning based classification of sheep behaviour from accelerometer data with imbalance Information Processing in Agriculture Sheep behaviour classification Data synthesis Class imbalance Grazing sheep |
| title | Deep learning based classification of sheep behaviour from accelerometer data with imbalance |
| title_full | Deep learning based classification of sheep behaviour from accelerometer data with imbalance |
| title_fullStr | Deep learning based classification of sheep behaviour from accelerometer data with imbalance |
| title_full_unstemmed | Deep learning based classification of sheep behaviour from accelerometer data with imbalance |
| title_short | Deep learning based classification of sheep behaviour from accelerometer data with imbalance |
| title_sort | deep learning based classification of sheep behaviour from accelerometer data with imbalance |
| topic | Sheep behaviour classification Data synthesis Class imbalance Grazing sheep |
| url | http://www.sciencedirect.com/science/article/pii/S2214317322000415 |
| work_keys_str_mv | AT kirketurner deeplearningbasedclassificationofsheepbehaviourfromaccelerometerdatawithimbalance AT andrewthompson deeplearningbasedclassificationofsheepbehaviourfromaccelerometerdatawithimbalance AT ianharris deeplearningbasedclassificationofsheepbehaviourfromaccelerometerdatawithimbalance AT markferguson deeplearningbasedclassificationofsheepbehaviourfromaccelerometerdatawithimbalance AT ferdoussohel deeplearningbasedclassificationofsheepbehaviourfromaccelerometerdatawithimbalance |