Jake Chen jakechen

Introduction

This quick guide describes how to create a Jupyter Notebook in AWS EC2 then how to access it remotely using SSH tunneling. This method is preferred since you do not open any additional ports besides 22, requires little-to-no configuration, and is generally more straight-forward.

Pre-requisites

This current version assumes basic familiarity with cloud computing, AWS services, and Jupyter Notebook. Mostly because this version won't have images and won't dive too deep into each individual step.

Steps

Spin-up EC2 instance with "Deep Learning" AMI

Log into EC2 console and click "Launch Instance" button.
Inside "AWS Marketplace", select the "Deep Learning AMI" from AWS. I use this AMI because most of the stuff you'll need is installed already.

	# Example uses GDELT dataset found here: https://aws.amazon.com/public-datasets/gdelt/
	# Column headers found here: http://gdeltproject.org/data/lookups/CSV.header.dailyupdates.txt

	# Load RDD
	lines = sc.textFile("s3://gdelt-open-data/events/2016*") # Loads 73,385,698 records from 2016
	# Split lines into columns; change split() argument depending on deliminiter e.g. '\t'
	parts = lines.map(lambda l: l.split('\t'))
	# Convert RDD into DataFrame
	from urllib import urlopen
	html = urlopen("http://gdeltproject.org/data/lookups/CSV.header.dailyupdates.txt").read().rstrip()

	import boto3
	import mxnet as mx
	from mxnet.io import NDArrayIter

	def predict_from_s3(record, bucket_name, s3_symbol_key, s3_params_key):
	"""Graphs MXNet network definitions from and S3 bucket and uses it for prediction on a single record

	Keyword arguments:
	record -- the record to predict from
	bucket_name -- bucket where your MXNet network is stored

	# With help from https://aws.amazon.com/blogs/ai/build-your-own-face-recognition-service-using-amazon-rekognition/

	frame_skip = 100 # analyze every 100 frames to cut down on Rekognition API calls

	import boto3
	import cv2
	from PIL import Image
	import io

	rekog = boto3.client('rekognition')