PSPNET - Plant and Soil Property Network
Abstract
PSPNet provides researchers easy access to machine learning tools in order to automate plant image analysis. As part of a new wildfire research project sponsored by the National Science Foundation [1], PSPNet aims to develop the foundation for a web platform to access machine learning tools. Using the web platform, researchers will be able to quickly classify their plant images using machine learning models so they may focus their attention on the wildfire research. Citizens are also welcomed to contribute plant images and collaborate with researchers to create comprehensive datasets.
Goal
PSPNet’s goal is to create a web application that provides image recognition tools to researchers studying wildfires. Ultimately, this reduces the amount of time professionals spend labeling images taken from the field, allowing solutions to combat wildfires to develop quickly.
Features
PSPNet allows users to upload images to identify using machine learning models. Upon uploading the images, the user can preview their images and add various information about the image dataset prior to submitting. Once the user's images are submitted and identified, the user can explore their and other user's image datasets. Viewing a single dataset reveals various information about the data, such as the various uploads from multiple users.
Future Work
Since PSPNet is only the foundation for the research project, there are several areas available for further development. PSPNet will have more user roles, including a principal investigator and lead investigator. PSPNet will offer a machine learning model which classifies plant and soil properties. Lastly, PSPNet will improve uploaded dataset functionality, including version control.
Conclusion
As part of a new wildfire research project, PSPNet sets the foundation for a web platform were plant images may be automatically labeled by machine learning models. Researchers and citizens can use PSPNet to collaborate and build comprehensive plant image datasets. With these automatically labeled datasets, researchers will be able to focus on discovering and researching the plants' effects on wildfires.
Resources
5-year research project involving DRI, UNLV, and UNR. The project "Harnessing the Data Revolution for Fire Science" aims to better understand wildfires in the sagebrush ecosystem.
YOLOv5 is an open-source machine learning model by Ultralytics. YOLOv5 incorporating lessons learned and best practices evolved over thousands of hours of research and development.
Pl@ntNet is a machine learning tool to help identify plants with images. This tool was a major inspiration and reference point for this project.