Summary

As part of the Open Innovation Pipeline, this project sought to trial a sophisticated visual recognition system, originally developed for other livestock industries, in livestock export facilities. The goal was to explore the use of artificial intelligence (AI) to automate cattle counting and weight estimation.

The trial aimed to replicate the conditions of a ship-loading environment, focusing specifically on a cattle race, to enhance the technology’s potential for implementation at livestock export loading and unloading sites. This approach considered practical constraints and risks associated with the process.

The AI system demonstrated 95% accuracy in weight estimation and 98% accuracy in cattle counting. Based on these results, the project provided recommendations for potential industry-wide adoption of the AI technology.

This project was managed by the Livestock Export RD&E Program, a collaboration between LiveCorp and Meat & Livestock Australia.

Objectives

• Design a comprehensive proof of concept (PoC) trial and methodology in collaboration with stakeholders, such as Registered Establishment (RE) operators and exporters.

• Develop indicative measures for evaluating trial success, in collaboration with industry stakeholders.

• Conduct trials and collected relevant data and information to form the final report.

Key findings

• CattleCounter had an overall accuracy of 98.97% in the trials, with improved significantly with site-specific training.

• CattleWeightEstimator had an overall accuracy of 90% in the trials, with the highest level of accuracy found when weighing animals between 300-500kg.

Benefits to industry

This technology has the potential to provide the livestock export industry with additional insights into the animals' wellbeing during voyages, as well as their performance. It also has applications for measuring weight change during time in pre-export quarantine yards and feedlots.