The main impetus for conducting this research was provide an easy way for ordinary citizens to participate in science as well as environment protection by using Machine Learning and Smartphones. As a first example, “Invasive Hydrangea Species Detection” was chosen because the invasive species are creating a significant problem, and data was readily available. Invasive Hydrangea Species cause major habitat loss for the native species and many other significant environmental detriments. Getting rid of these plants are costly due to specialists having to manually track them down and getting rid of each one individually. Our goal is to solve this by allowing ordinary citizens to identify invasive species with ease.
We got our data from the Kaggle competition Invasive Species Monitoring. It consisted of the different labeled images of hydrangea species that acted as “training” in classifying the species. In brief, we classified 2,000 pictures for training and the rest was used to do the actual classification and test the accuracy of our machine. With different Machine Learning algorithms, we were able to increase the accuracy of our classification machine. At last, we consolidated all our findings into a single phone app that allowed the user to identify whether a plant was invasive or native and report it to the authorities by using only the default camera features.
The machine learning resulted in the highest accuracy of 96.27% accuracy, meaning that in most cases any invasive species will be efficiently identified by our program. The smartphone app with the same accuracy was also created, ready to be distributed freely.