Latest News

Good Clover Bad Clover Project
MFMG has secured funding from MLA for a ‘Good Clover, Bad Clover’ project which will focus on upskilling producers on oestrogenic clover identification and the implementation of practices to improve lamb marking...

Read More


Canola Growers Urged to Check Revised Blackleg Ratings

Grain growers in the southern cropping region are being urged to check the latest blackleg disease ratings for canola cultivars ahead of sowing in 2017. 

 

Read More


After a good crop in 2016, what does it take to restore soil fertility?

This article uses data from the Bool Lagoon and other sites, on crop nutrient replacement after the 2016 season

 

Read More


Read All News

Automated Condition Scoring Project

 

We plan to develop an automated condition scoring system for sheep that will allow objective body condition measures to be taken immediately off-shears. The development of automated, objective body condition scoring systems will reduce the subjectivity associated with manual condition scoring, and significantly decrease the time and labour inputs required to accurately condition score mobs. 

 

This project will capture images of freshly shorn sheep (as this is the easiest time to “see” condition score) with a hyperspectral camera to produce infrared images.  Using these images and the actual condition score and liveweight, machine learning will be used to automatically predict condition score of sheep.

 

There are three parts to the project.

1. Prototype development:  As the height, angle and location of the camera affects the images, it is necessary to optimise the camera location.  During the prototype development we will collect multiple images from different locations around the sheep to determine optimum positioning of the camera and other image metrics such as pixel density and number of images required per sheep.

2. Machine learning: Up to 1000 images need to be captured to enable the machine to “learn” what different condition scores “look” like.  These images need to cover the breadth of actual condition scores across different sheep.  We will visit 10 to 15 properties at shearing and weigh, condition score and capture images of approximately 100 sheep per property.

3. Validation:  Once the machine has “learnt” how to condition score, we need to test that it actually knows what it is doing.  We will therefore visit up to 10 new properties at shearing to test that the machine can actually predict the condition score of the sheep.  These images will also be used to further “train” the machine.

This is part of a larger project in the Sheep CRC and the overall hypothesis behind the research program is that better use of information on body condition, weight change, genetic background and previous production history can be used to improve both wellbeing and productivity.

 

Anyone that is interested in more information, or if you are shearing in late October, November, December ‘16 or January ‘17 and would like to be involved in stage 2 (machine learning), please contact:

 

Emma Babiszewski

emma.babiszewski@sa.gov.au

Mob: 0427 000 264

Ph: (08) 8762 9185