Automated Individual Pig Localisation, Tracking and Behaviour Metric Extraction Using Deep Learning

Jake Cowton,Ilias Kyriazakis,Jaume Bacardit

doi:10.1109/access.2019.2933060

Jake Cowton, Ilias Kyriazakis + Show 1 more

Open Access

https://doi.org/10.1109/access.2019.2933060

Copy DOI

Abstract

Individual pig tracking is key to stepping away from group-level treatment and towards individual pig care. By doing so we can monitor individual pig behaviour changes over time and use these as indicators of health and well-being, which, in turn, will assist in the early detection of disease allowing for earlier and more effective intervention. However, it is a much more computationally challenging than performing this task at group level; mistakes in identification and tracking accumulate and, over time, provide noise measures. We combine a deep CNN object localisation method, Faster Region-based convolutional neural network (R-CNN), with two potential real-time multi-object tracking methods in order to create a complete system that can autonomously localise and track individual pigs allowing for the extraction of metrics pertaining to individual pig behaviours from RGB cameras. We evaluate two different transfer learning strategies to adapt Faster R-CNN to our pig detection dataset that is more challenging than conventional tracking benchmark datasets. We are able to localise pigs in individual frames with 0.901 mean average precision (mAP), which then allows us to track individual pigs across video footage with 92% Multi-Object Tracking Accuracy (MOTA) and 73.4% Identity F1-Score (IDF1), and re-identify them after occlusions and dropped frames with 0.862 mAP (0.788 Rank 1 cumulative matching characteristic (CMC)). From these tracks we extract individual behavioural metrics for total distance travelled, time spent idle, and average speed with less than 0.015 mean squared error (MSE) for each. Changes in all these behavioural metrics have value in the detection of pig health and wellbeing.

Highlights

The ability to monitor the behaviour of animals, in particular, how said behaviour changes over time and under varying circumstances, provides us with knowledge that can assist in identifying problems before they become serious, or even life threatening, and enhances the success of intervention [1], [2]
MATERIALS & METHODS we describe the data used to train and evaluate our methods for detection, tracking and extracting behavioural metrics, the model used to detect the location of pigs in an image, the tracking methods that were used to assign the detections to identities, and how we analysed the tracks to extract behavioural metrics
As Multi-Object Tracking Accuracy (MOTA) has shortcomings in terms of how it accounts for identity switches [26], we report tracking metrics using Identity F1-Score (IDF1), which is included as a metric in the multi-object tracking (MOT) Challenge, as this provides a much more global way to assess the performance of the tracking system in terms of its ability to track identities

Summary

Introduction

The ability to monitor the behaviour of animals, in particular, how said behaviour changes over time and under varying circumstances, provides us with knowledge that can assist in identifying problems before they become serious, or even life threatening, and enhances the success of intervention [1], [2]. Able to detect substantial changes or clinical signs They fail to detect subtle changes in behaviour that usually precede clinical signs of disease; this results in late intervention [1], [3]. We have previously developed a method that enables us to measure behavioural traits in groups of pigs [3]. Despite the usefulness of having group-level pig behaviour measures, there are several advantages in being able to automatically detect individual behaviours and identify pigs that may be at risk or are challenged [2] as this enables more personalised treatment plans.

Methods

Results

Discussion

Conclusion