Deploying a Sustainable Deep Learning Pipeline for Poison Ivy Image Classification

Poison ivy, causing irritated skin in humans due to the urushiol oil it contains, is a well-known plant mainly distinguishable by the traits of its leaves. Nevertheless, according to the statistical data, a significant number of people are still affected by the plant. This research postulates that this notorious result stemmed from the scarce analysis about the plant. The paper proposes an automated pipeline, functioning Continuous Integration, Deployment, and Training on Kubeflow so that the harmful impact of the plant decreases, leading the public health outcomes to enhance. The pipeline contains a dataset composed of both a set of leaves and a stem images which are manually collected from a local farm and a devised model which considers both features to classify Western poison ivy. With the pipeline, the research contributes to decrease the number of suffering people from the plant as promptly reacting to seasonal changes.