Identifying Rocky Intertidal Zone Plant Species using Convolutional Neural Networks
Ecological monitoring of plant and animal species helps in maintaining ecological balance. It helps in understanding the species, their assemblage, changes that occur in their assemblage and factors causing those changes. The present study involves monitoring of plant species in rocky intertidal zones of Santa Rosa Island, California. Traditionally, ecological monitoring has been done using photo transects. These photo transects are then quantified manually by humans, but quantifying a huge amount of data manually can be time-consuming and prone to errors. The present study helps to address these problems by using a machine learning technique - semantic segmentation. Additionally, image classification is also performed. The study involves building two convolutional neural networks - one from scratch and the other using transfer learning on a publicly available network. Datasets used in the study were collected by the Biology department at California State University, Channel Islands and the network is built using a publicly available framework - Keras.