Automating Animal Monitoring & Staff Behavior Detection

for a Leading Network of Dairy Farms

#Data Science & AI
Get free consultation

Duration

The project was implemented over
3
months

Client

Health & Fitness Technology
US
based

Project Overview

The client needed a solution to continuously monitor calf activity levels to assess animal health and to detect any incidents of animal mistreatment by identifying signs of aggressive behaviour from farm staff. This required an advanced system capable of tracking the animals’ behaviour in real time, identifying early signs of health issues, and flagging any aggressive actions by staff members to improve both operational efficiency and animal welfare standards. The solution needed to integrate seamlessly into existing operations while providing accurate insights for both animal health monitoring and human behaviour analysis.

Challenge

First, accurately tracking calf activity levels to monitor their health was complex, requiring a system capable of detecting subtle changes in behavior, such as standing, lying, or inactivity, which could indicate early signs of illness. 
Second, detecting and preventing aggressive behavior from farm staff toward the animals was crucial. This required a reliable method to identify aggressive interactions, including enhancements for recognizing individuals despite challenges such as uniform clothing or distance from the camera.

Provided Services

AI/ML Development
Real-time Analytics
Data Science
System Integration
Custom Model Training
Computer Vision

Tools and Technologies

Solution

Stellans implemented a comprehensive solution comprising two main components:

1

Animal Activity Monitoring:

Compiled a dataset and trained an object detection model using active learning to detect cows in standing and lying positions and track their activity levels.

2

Aggressive Behaviour Detection: Two-Stage Approach

First, farm staff were detected and skeletal keypoints were extracted to enable tracking, though uniform clothing initially caused difficulties. To solve this, object and keypoint detection models were retrained using client-specific annotated data. In the second stage, keypoints from each staff track were used to train a transformer-based neural network that identified aggressive behaviour.

Results

faster response time

15
%
faster response time to health problems compared to manual methods.

Reduced Risk of Animal Abuse

70
%
more cases of staff aggression toward animals identified , reducing abuse risks and enhancing the farm’s humane reputation.

TESTIMONIALS

Ethan Grundleger

Ethan Grundleger

Director Business Strategy & Intelligence

Ethan Grundleger company

"It’s impressive how dedicated the employees are."

Verified Verified review
Matt Miritello

Matt Miritello

VP of Engineering

Matt Miritello company

"Their depth of knowledge made all the difference."

Verified Verified review
Slava Kononenko

Slava Kononenko

CMO/CPO

Slava Kononenko company

"They accomplished everything we wanted to achieve."

Verified Verified review
Inna Igontova

Inna Igontova

CMO / Co-founder

Inna Igontova company

"Stellans’ expertise has really helped us sharpen our marketing and product strategies."

Verified Verified review
Jeremiah Zinn

Jeremiah Zinn

Chief Product Officer

Jeremiah Zinn company

"Their impact was transformative and enduring."

Verified Verified review
Roman Degtyur

Roman Degtyur

Chief Information Officer

Roman Degtyur company

"They didn’t just solve problems, they redefined what excellence looks like in data operations."

Verified Verified review
Konstantin Luchyna

Konstantin Luchyna

CMO

Konstantin Luchyna company

"Stellans quickly became our go-to data partner. Smart, fast, and easy to work with."

Verified Verified review

Related case studies

    Get a Free Consultation

    Let's talk about
    your project

    * You can attach up to 3 files, each up to 3MB, in doc, docx, pdf, ppt, or pptx format.