Project collaborators
Jeri Barak and Kim Cowles, Plant Pathology
Arjun Iyer and Iain McConnell, Data Science Institute
Project start and end dates
12/1/2022-5/1/2023
Project summary
“Every time, walking back from DSI, Kim and I were energized and excited about doing science.” —Jeri Barak
Jeri and Kim wanted to automate the data collection process of their experiments: segmenting images of leaves into healthy, lesioned, and not-leaf areas. There was very little preexisting data, as each manually annotated leaf image took a comparatively long time to produce using existing published methods.
Given the desire for a quick turnaround and the constrained data collection, DSI scoped the work as a non-AI image processing application, hired undergraduate student Arjun to develop the application, and provided Data Scientist Iain to supervise that student for the duration of the project. The main output of Arjun’s work was an image processing application hosted online. This tool can be used by Kim and Jeri, or anyone else with appropriate leaf images.
To achieve a usable result, given the short window for a result and the limited available datasets, the plant pathology image processor does not use any machine learning models. Instead, it uses a deterministic image processing pipeline based on sklearn-image to achieve its image segmentation results.
The immediate impact of the work was to speed up Kim and Jeri’s data collection process by automating a time-consuming, manual process. Further, Arjun received undergraduate research credits from working in the Barak lab. This work also paved the way for ongoing collaboration between DSI and the Barak lab.
Project outputs/deliverables
- Web app
- CALS Data Science Showcase poster (produced by Kim Cowles)