Time Series Classification to Improve Poultry Welfare

Abstract

Poultry farms are an important contributor to the human food chain. Worldwide, humankind keeps an enormous number of domesticated birds (e.g. chickens) for their eggs and their meat, providing rich sources of low-fat protein. However, around the world, there have been growing concerns about the quality of life for the livestock in poultry farms; and increasingly vocal demands for improved standards of animal welfare. Recent advances in sensing technologies and machine learning allow the possibility of automatically assessing the health of some individual birds, and employing the lessons learned to improve the welfare for all birds. This task superficially appears to be easy, given the dramatic progress in recent years in classifying human behaviors, and given that human behaviors are presumably more complex. However, as we shall demonstrate, classifying chicken behaviors poses several unique challenges, chief among which is creating a generalizable "dictionary" of behaviors from sparse and noisy data. In this work we introduce a novel time series dictionary learning algorithm that can robustly learn from weakly labeled data sources.

Features or Shape? Tackling the False Dichotomy of Time Series Classification

“…These exclusion window lengths can be set by common sense (especially for human activities), learned from strongly labeled data, or set using domain knowledge. For example, for the chicken dataset we describe later, the literature suggests that two examples of pecking should be at least ¼ of a second apart [15] [21].…”

“…Pecking is the act of striking the ground with beak, while dustbathing is the act of grooming by rolling in the dirt. We used the dataset introduced into the community by [15]. The training data as illustrated in Figure 9 is a 24-hour onedimensional time series of length 8,658,539 datapoints (Blue) with its corresponding annotations (Red).…”

“…The peck behavior has highly conserved shape. While the dustbathing behavior does not have a stereotypical shape, it is possible to recognize by several features, including "complexity" [15][23] [28]. Moreover, we believe that many problems have this property.…”

Time Series Classification: Lessons Learned in the (Literal) Field while Studying Chicken Behavior

“…We are not proposing that all chickens be monitored, that is clearly unfeasible. Our system is designed as a tool to allow researchers to assess the effects of various conditions on chicken health, and then use the lessons learned on the entire brood [4].…”

“…The classification of chicken behaviors throughout the day has been discussed in [4]; in which computer scientists extracted a dictionary of well-preserved chicken behaviors (i.e. pecking, preening and dustbathing) from a short video annotated chicken dataset, as shown in Figure 5.…”