Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Moo-ving Beyond Tradition: Revolutionizing Cattle Behavioural Phenotyping with Pose Estimation Techniques

Navid Ghassemi, Ali Goldani, Ian Q. Whishaw, Majid H. Mohajerani

TL;DR

The paper surveys how pose estimation can transform cattle behavioural phenotyping within precision livestock farming, framing data quality, model design, and evaluation as critical factors for real-world impact. It analyzes current DL architectures, data practices, and metrics, and identifies gaps in data diversity and standardization for crowded farm environments. A key contribution is the proposal of Open Cattle Hub to democratize data, unify labeling and benchmarks, and bridge industry with academia, enabling scalable development of cattle pose-estimation solutions. The work highlights a path toward modular, open systems that can improve health monitoring, welfare assessment, and productivity across diverse farm settings, with practical implications for adoption and policy in the cattle industry.

Abstract

The cattle industry has been a major contributor to the economy of many countries, including the US and Canada. The integration of Artificial Intelligence (AI) has revolutionized this sector, mirroring its transformative impact across all industries by enabling scalable and automated monitoring and intervention practices. AI has also introduced tools and methods that automate many tasks previously performed by human labor with the help of computer vision, including health inspections. Among these methods, pose estimation has a special place; pose estimation is the process of finding the position of joints in an image of animals. Analyzing the pose of animal subjects enables precise identification and tracking of the animal's movement and the movements of its body parts. By summarizing the video and imagery data into movement and joint location using pose estimation and then analyzing this information, we can address the scalability challenge in cattle management, focusing on health monitoring, behavioural phenotyping and welfare concerns. Our study reviews recent advancements in pose estimation methodologies, their applicability in improving the cattle industry, existing challenges, and gaps in this field. Furthermore, we propose an initiative to enhance open science frameworks within this field of study by launching a platform designed to connect industry and academia.

Moo-ving Beyond Tradition: Revolutionizing Cattle Behavioural Phenotyping with Pose Estimation Techniques

TL;DR

The paper surveys how pose estimation can transform cattle behavioural phenotyping within precision livestock farming, framing data quality, model design, and evaluation as critical factors for real-world impact. It analyzes current DL architectures, data practices, and metrics, and identifies gaps in data diversity and standardization for crowded farm environments. A key contribution is the proposal of Open Cattle Hub to democratize data, unify labeling and benchmarks, and bridge industry with academia, enabling scalable development of cattle pose-estimation solutions. The work highlights a path toward modular, open systems that can improve health monitoring, welfare assessment, and productivity across diverse farm settings, with practical implications for adoption and policy in the cattle industry.

Abstract

The cattle industry has been a major contributor to the economy of many countries, including the US and Canada. The integration of Artificial Intelligence (AI) has revolutionized this sector, mirroring its transformative impact across all industries by enabling scalable and automated monitoring and intervention practices. AI has also introduced tools and methods that automate many tasks previously performed by human labor with the help of computer vision, including health inspections. Among these methods, pose estimation has a special place; pose estimation is the process of finding the position of joints in an image of animals. Analyzing the pose of animal subjects enables precise identification and tracking of the animal's movement and the movements of its body parts. By summarizing the video and imagery data into movement and joint location using pose estimation and then analyzing this information, we can address the scalability challenge in cattle management, focusing on health monitoring, behavioural phenotyping and welfare concerns. Our study reviews recent advancements in pose estimation methodologies, their applicability in improving the cattle industry, existing challenges, and gaps in this field. Furthermore, we propose an initiative to enhance open science frameworks within this field of study by launching a platform designed to connect industry and academia.
Paper Structure (21 sections, 2 figures, 1 table)

This paper contains 21 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Development Process for Pose Estimation Methods
  3. Data, the fuel of proper model training
  4. How to access all the needed data?
  5. Synthetic Data, a Workaround Solution
  6. Training a deep learning model
  7. Model Component
  8. Pre-Processing
  9. Backbone
  10. Neck
  11. Head
  12. Handling Multi-instances
  13. Segmenting Body Parts
  14. State of the art methods
  15. Selecting Best Model
  16. ...and 6 more sections

Figures (2)

  • Figure 3: Pose estimation can provide information regarding animals' behaviour (e.g., eating vs walking), specific events (e.g., calving), genetics (e.g., breed), animal age, and health conditions. Top Left: Eating behaviour, an activity that can be found in data using posture, and its variations through time. Top Right: Calving event, approximation of which can be estimated using posture changes through time speroni2018increasing. Bottom: Lameness, detected by observing the bovine back arc sprecher1997lamenessscott1989changes.
  • Figure 4: Many parameters can cause a change in distribution of data for cattle behavioural phenotyping, some of which are shown in this figure. Camera View: By changing the view of camera, not only the data would change, but also some keypoints might become invisible. Lighting: The time of day and the lighting of the environment have a visible effect on data. Environment: Occlusions to view, or extreme weather conditions can come with different environments, and thus change characteristics of data. Crowdedness: Another parameter that introduces occlusions to the view and shifts to distribution is subjects overlapping in the captured image, courtesy of crowdedness in data.