Location: Sustainable Agricultural Systems Laboratory
Project Number: 8042-22000-167-120-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Sep 30, 2024
End Date: Sep 29, 2029
Objective:
Create a Digital Ag Science Hub (DASH) to translate the ARS’s expertise in breeding and production research into highly curated deployable solutions.
Specifically we will:
1. Expand SCINet compute resources to include virtual machine servers that support databases of meta data for images and graphical user interfaces.
2. Construct machine learning operations that use continuous integration/continuous deployment pipelines
3. Build image repositories on SCINet
4. Containerize models and publish on docker hub.
5. Models deployed onto edge devices, personal computers, cloud, and SCINET virtual machines.
6. Provide education and training on use of computer vision and artificial intelligence tools, phenotyping solutions with satellites, low-cost cameras, and drones
Approach:
Key implementation features of DASH.
Enterprise-level deployable solutions will use the following framework:
1. Target identification: Implement a universal intake process using a scoping document that identifies likely phenotyping targets with a researcher and any barriers to success.
2. Sensor/ platform selection: Standardize technology choices to promote easier collaboration among researchers; curate a suite of sensors and platforms; develop decision key for informing sensor selection and platform.
3. Annotation: Curate open-source annotation/labeling tools available for each style of annotation and make them easily installable on the ARS computers and SCINet.
4. ML Algorithm development: Provide a modern continuous integration/continuous deployment (CI/CD) pipeline so that anyone training an ML algorithm has version control, offsite backup, and automated containerization of code.
5. Deployment: Create a tool for deploying containerized ML algorithms onto SciNet, cloud, and edge devices.
6. ML improvement: Define feedback loops for continued improvement of ML algorithms.
7. Education and training: Make available online tutorials, user workshops, office hours, and hotline support.
