Source: AGRINERDS, INC. submitted to
CONNECTING THE DOTS: LINKING A NOVEL EGG COUNTING DEVICE TO MACHINE LEARNING BASED SOFTWARE TO FACILITATE IMPROVED FOOD SAFETY AND PRODUCTION EFFICIENCY IN POULTRY PRODUCTION
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
SMALL BUSINESS GRANT
Reporting Frequency
Annual
Accession No.
1025731
Grant No.
2021-33530-34399
Cumulative Award Amt.
$100,000.00
Proposal No.
2021-00996
Multistate No.
(N/A)
Project Start Date
Jul 1, 2021
Project End Date
Aug 31, 2022
Grant Year
2021
Program Code
[8.3]- Animal Production & Protection
Project Director
Pitesky, M.
Recipient Organization
AGRINERDS, INC.
2208 HUMBOLDT AVE
DAVIS,CA 956163086
Performing Department
Pop Health & Reproduction
Non Technical Summary
One of the primary challenges of egg counting and data analysis in poultry production is the disparate nature of poultry production. Different husbandry systems (e.g. caged, organic, aviary) create different types of challenges. For example, based on our interaction with largelayer companies through our commercial client Hy-Line International, the ability to count eggs with the current technology (i.e. Big Dutchman belt counters) is highly inaccurate due to the increased width of egg belts.Hence improvements need to prioritize flexibility for how different poultry producers can utilize hardware and software. In addition, it is important to recognize that more data is collected now than ever before. The ability to automatically collect new streams of data such as vocalization and image analysis along with more granular versions of previously collected data such as light intensity, air quality, water consumption and temperature portend a new opportunity for prediction and causal analyses of retrospective data. Hence our proposed innovation to more accurately quantify laying production at a row level within a barn and linking those data to other already collected data such as light intensity and feed and water consumption, can be used to develop novel predictive insights associated with productivity and profitability.In addition, the ability to use and link the ML based predictions with decision science tools like Analytical Hierarchy Processes (AHP) would allow companies to better integrate predictive ML based statistics with institutional knowledge at the company level. As we further explore both approaches it is important to not rely too heavily on either approach. Therefore, our software would allow decision makers at the company level to see alignment or misalignment between the ML based output and the decision science based output reflecting stakeholder opinion.
Animal Health Component
100%
Research Effort Categories
Basic
(N/A)
Applied
100%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
30732991170100%
Knowledge Area
307 - Animal Management Systems;

Subject Of Investigation
3299 - Poultry, general/other;

Field Of Science
1170 - Epidemiology;
Goals / Objectives
Demonstrate the technical feasibility of developing a novel egg counting devicethat will interface with current commercial software and/or our web-based software.Develop integrative software that utilizes optimized machine learning (ML) and decision science based tools to improve production efficiency, poultry health and food-safety along the entire poultry production supply chain.
Project Methods
Methods of data analysis will require exploring various ML based approaches with data provided by our commercial partners to understand the predictive ability of each ML based approach. ML based approaches that will be tested will include: Random Forest, Decision Trees, Boosting. Before this can be accomplished we will develop tools to accurately ingest raw production data into a cloud based relational databaseWith respect to the egg counter, we willconduct simulated testing and resolve hardware bugs, then manufacture the separator devices for accurate counting and assemble the finished unit, and finally field test the device at a partner farm (JS West in Modesto, CA). At this point additional testing will be done in order to gauge its performance as compared to their current egg counting methods. In the sensor testing phase, we will use an oscilloscope, power supply and signal generator instruments oto provide more stable input signals and measure circuit outputs. Once we have a complete circuit design that works in a prototype setting and is verified through simulation software, we will design and manufacture printed circuit boards (PCBs) for assembly and integration into the final prototype device. In tandem with this PCB design process, we will continue design and printing of the egg separators for testing at JS West.

Progress 07/01/21 to 08/31/22

Outputs
Target Audience:Our target audience are commercial poultry producers (turkey, broiler and layer) that typically have multiple barns/farms. These facilities typically (based on our work with a large genetics company) have the greatest deviation in egg counts between numbers automatically counted in the barn and the numbers of eggs counted in the processing facility. On average we have identified a5% deviation in egg counts using the current standard overhead infrared (IR) counterswith miscounts up to 15% (over and under the actual number realized in the egg processing plant). Changes/Problems:We have pivoted the software to focus more on economic/productivityanalysis as opposed to just production and food safety based on feedback from potential customers. What opportunities for training and professional development has the project provided?We were able to send one of our employees to the Poultry Tech Summit to present our data on the egg counting machine learning algorithim. We have also been able to work with several commercial partners which has provided a unique opportuntity for our engineers to work in the agricultural space. As they don't have expertise in poultry, this has provided valuable training opportuntiies to understand additional challengs in the poultry space that AgriNerds can provide valuable technological solutions for. How have the results been disseminated to communities of interest?Results have been disseminated via every other month updates with the commercial layer company and bi-monthly meeting with the commercial turkey producer. The egg counter technology was presented at the 2022 Poultry Tech Summit in Atlanta, GA What do you plan to do during the next reporting period to accomplish the goals?Deployment of the egg counter for intiial trial runs and deployment of the software for beta testing. We are currently installing the egg counters into one of the top 10 largest egg producers in the U.S. We are nearing completion of scalable software for data analysis of production related data for the turkey industry which we anticipate will be scalable to the layer and broiler industry.

Impacts
What was accomplished under these goals? Objective: Demonstrate the technical feasibility of developing a novel egg counting device that will interface with current commercial software and/or our web-based software. Using a custom-trained image classification and object tracking tool, we are currently able to count eggs with at least a 99% accuracy. The system has been tested at multiple stages in the egg collection process for two different companies in order to demonstrate the ability of the image classifier to create novel image classificationmodels for each facility. Basic strategies have been developed to determine the size of the eggs detected and to assess the eggs for defects, but additional testing will be required to develop these strategies to be robust enough to be used in production. In summary, we have developed an initial ML based algorithim for egg counting and working on submitting a ROI (Record of invention) for the egg counting device.While the performance of the system thus far has been better than the current IR based technology,it will be important to continue to assess the accuracy to ensure that neither debris accumulation nor other factors diminish the performance over time. Objective: Develop integrative software that utilizes optimized machine learning (ML) and decision science based tools to improve production efficiency, poultry health and food-safety along the entire poultry production supply chain. standardize the ingestion of the data using several custom Extract-Transform and Load (ETL) protocols representing over 10 commercial layer companies across the U.S. Development of an automated method of validation through conditional formatting on master excel files which canc interface withthrough their custom software which for our one commercial client is then able to share with their clients. Our commercial client has multiple clients so we have a system set up where we charge by the company to the "enterprise client." We are currently working on similar software for a large commercial turkey company and anticipate having a commercial scalableversion market ready in the next 3-4 months. We anticipate presenting this at the 2023 Poultry Tech Summit in the Fall.

Publications


    Progress 07/01/21 to 03/14/22

    Outputs
    Target Audience:commercial poultry producers (turkey, broiler and layer). Changes/Problems:We have pivoted the software to focus more on economic analysis as opposed to just production and food safety based on feedback from potential customers. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?Results have been disseminated via every other month updates with the commercial layer company and bi-monthly meeting with the commercial turkey producer. What do you plan to do during the next reporting period to accomplish the goals?Deployment of the egg counter for intiial trial runs and deployment of the software for beta testing.

    Impacts
    What was accomplished under these goals? Demonstrate the technical feasibility of developing a novel egg counting device that will interface with current commercial software and/or our web-based software. We have developed an initial ML based algorithim for egg counting and working on submitting a ROI (Record of invention) for the egg counting device. Develop integrative software that utilizes optimized machine learning (ML) and decision science based tools to improve production efficiency, poultry health and food-safety along the entire poultry production supply chain. The software is being developed in collaboration with a large mid-western turkey company. We anticipate having the first version depolyed in the next 3-4 months,

    Publications


      Progress 07/01/21 to 02/28/22

      Outputs
      Target Audience:Commercial poultry producers. We are currently working with an international poultry genetics company, a large midwestern turkey producer and a large west coast layer company. Changes/Problems:We have shifted away from machine learning (ML) for our software to focus on data integration. We anticipate using a ML based approach as the amount of data we capture increases sufficiently. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?Yes we are working in concert with several poultry companies and have frequent updates with them on our progress. What do you plan to do during the next reporting period to accomplish the goals?Test the egg counter in a layer house and deploy the first version of our software with the large midwestern poultry company.

      Impacts
      What was accomplished under these goals? With respect to our egg counter we are in the process of submitting an ROI (record of invention). We have also started our initial machine learning (ML) for our egg counting system. With respect to our software we are working witha large mid-western poultry company on developing novel sofware that dovetails multiple data sources to optimize economic predictions. We anticipate have the alpha version completed by the late spring.

      Publications