CDI DL WORKSHOP

A collection of jupyter notebooks for workshops on image classification and recognition using deep transfer learning

Written by Dr Daniel Buscombe Northern Arizona University [email protected]

This workshop was prepared for the "MAPPING LAND-USE, HAZARD VULNERABILITY AND HABITAT SUITABILITY USING DEEP NEURAL NETWORKS" project, funded by the U.S. Geological Survey Community for Data Integration, 2018

Thanks: Jenna Brown, Paul Grams, Leslie Hsu, Andy Ritchie, Chris Sherwood, Rich Signell, Jon Warrick

These materials and instructions are currently for workshop participants only, and will be updated in October for general use

Many of these materials are mirrored in the dl_tools library

General Instructions

Clone repo

git clone https://github.com/dbuscombe-usgs/cdi_dl_workshop.git

cd cdi_dl_workshop

Create conda environment

conda env create -f binder\environment.yml

conda activate cdi_workshop

Launch jupyter

python -m ipykernel install --user --name cdi_workshop --display-name "Python (cdi)"

jupyter notebook

Workshop Instructions

Fork this repository

<img width="960"

Launch

ESIPFED jupyter cloud computer

Go to http://pangeo.esipfed.org and log in with your github credentials. Note, this will only work if you are a member of the cdi-workshops github group (invitation only)

When your server starts up, you should see a jupyter in traditional 'tree' view

You can work in newer 'lab' view instead by modifying the url from 'tree' to 'lab'

Clone the repo

Clone YOUR forked repository from YOUR github page

Launch a notebook