Machine Learning for Solid Waste Management

Install Anaconda Python 3:

• https://www.anaconda.com/download
• https://www.digitalocean.com/community/tutorials/how-to-install-the-anaconda-python-distribution-on-ubuntu-16-04

Install Needed Extra Libraries:

• pip install --upgrade google-api-python-client
• pip install boto3
• pip install awscli --upgrade --user

Install Docker and the caffe docker image:

• sudo apt-get install docker.io
• docker run -ti bvlc/caffe:cpu caffe --version

Make sure your AWS credentials are setup:

Setup ~/.aws/credentials with your key and secret:

[default]
aws_access_key_id = YOUR_KEY
aws_secret_access_key = YOUR_SECRET

Set the default region in ~/.aws/config:

[default]
region=ap-south-1

Updating the dataset

1. Export the Google Drive Spreadsheet as a csv file
2. Run ./process-csv csv-file-name
3. Run ./package-dataset output-folder-path

Use the flag -f to force syncing with Google drive otherwise cached local copies will be used if available.

Downloading and Packaging the Dataset

1. Run ./package-dataset directory-to-store-tarball

This will create a folder in your home directory called 'swm-ml-dataset' and save a tarball to the folder you specify.

Extracting Caffe Features

1. sudo docker run -ti -v ~/swm-ml-dataset/images:/opt/caffe/swm-ml-dataset -v /home/ubuntu/swm-ml:/opt/caffe/swm-ml bvlc/caffe:cpu /bin/bash

Authors

• Andrew Ziegler - andrewzieg@gmail.com