This repo provides a framework for a machine learning (ML) model that can be trained to classify images of phytolankton into different groups. The demo uses 5,000 images of phytoplankton, with 1,000 each of five different groups. It includes a self-contained Jupyter notebook that can be run in Google Colab, a descriptive report of the project, a document containing relevant resources, and a zipped folder of data needed to run the demo.
The intent of this repository is to allow users to test and experiment with training a Convolutional Neural Network (CNN), which is a specific type of deep neural network especially useful for image classification. The demo can be run by anyone, whether or not the user has any prior experience with python or deep learning. We hope you find it useful!!
This Jupyter notebook has a link at the top to run it in Google Colab. If it is run in Colab, no installation of python is required.
This report contains a description of the motivation, methods, results, and some discussion and reflection on the use of CNNs for plankton imagery classification.
Author: Hisham Bhatti, University of Washington, 2023
This document contains links to a variety of resources on topics relevant to this demo project. Topics include collection of plankton imagery data, basic machine learning concepts & convolutional neural networks, and python.
Authors: Hisham Bhatti & Ali Chase, University of Washington, 2023
This is a zipped folder containing five subfolders of plankton images. Each of the five folders contains 1,000 images of a given group of phytoplankton. The five groups are: Chaetoceros (a type of diatom), Cryptophytes, Dinoflagellates, Euglenoids, and Oxytoxum (a type of dinoflagellate). Size unzipped: 28.1 MB